Demonstration of an azobenzene derivative based solar thermal energy storage system †

Molecules capable of reversible storage of solar energy have recently attracted increasing interest, and are often referred to as molecular solar thermal energy storage (MOST) systems. Azobenzene derivatives have great potential as an active MOST candidate. However, an operating lab scale experiment including solar energy capture/storage and release has still not been demonstrated. In the present work, a liquid azobenzene derivative is tested comprehensively for this purpose. The system features several attractive properties, such as a long energy storage half-life (40 h) at room temperature, as well as an excellent robustness demonstrated by optically charging and dis-charging the molecule over 203 cycles without any sign of degradation (total operation time of 23 h). Successful measurements of solar energy storage under simulated sunlight in a micro ﬂ uidic chip device have been achieved. The identi ﬁ cation of two heterogeneous catalyst systems during testing enabled the construction of a ﬁ xed bed ﬂ ow reactor demonstrating catalyzed back-conversion from cis to trans azobenzene at room temperature under ﬂ ow conditions. The working mechanism of the more suitable catalytic candidate was rationalized by detailed density functional theory (DFT) calculations. Thus, this work provides detailed insights into the azobenzene based MOST candidate and identi ﬁ es where the system has to be improved for future solar energy storage applications.


Introduction
The development of sustainable energy has been attracting increasing attention, due to the pressing environmental and social challenges linked to high dependency on fossil fuels in our modern society. 1 As the most abundant energy source for Earth, the Sun can provide the energy needed for mankind for an entire year in only six hours. 2 Various ways to take advantage of solar energy have been studied and developed over the last few decades, including photovoltaics, 3 articial photosynthesis, 4 and solar thermal heating. 5 In order to store sunlight for future use, molecular solar thermal storage techniques (MOST, 6,7 also known as solar thermal fuels, STF 8 ) focus on harvesting solar energy and storing it in photoswitchable materials. A parent molecule can be isomerized by solar irradiation to a metastable high energy photoisomer for a long storage period. When the energy is required, it can be thermally or catalytically back-converted to the parent state. Ideally, the stored energy should be released on demand as heat, thus operating as a closed-cycle system.
In order to realize the MOST concept, several features concerning the charge and discharge process need to be considered. 7,9,10 (1) Since more than 50% of sunlight is distributed from 300 nm to 800 nm, the parent molecules should be able to absorb broadly in this spectral region. (2) The quantum yield of the photoisomerisation reaction should be as close to unity as possible. (3) The storage half-life at room temperature should be long enough to fulll application-relevant storage times, such as daily, monthly, yearly or even longer storage requirements. (4) The energy of the metastable photoisomer should be signicantly higher than the ground state of the parent isomer. (5) The system should ideally be able to operate over an innite number of charging and heat releasing cycles. (6) Using a heterogeneous catalyst, the photoisomer should easily release the stored energy as heat. (7) For energy collection, storage, and bulk heating applications, MOST materials need to be pumped between solar collector, storage reservoir and heat extraction devices. As a consequence, the MOST system should be a liquid or highly soluble solid.

Results and discussion
To characterize the physical properties of AZO1, it was rst studied in toluene in terms of its absorptivity, half-life of energy storage and quantum yields of the photoisomerization reaction. Comparing these results with available data of AZO1 in methanol, 43 the maximum in the absorption spectrum was observed to be slightly red shied, indicating a low solvent polarity effect compared to, for instance, a DHA-MOST system 18 (ca. 5 nm, 3 Max@350 nm ¼ 2.6 Â 10 4 M À1 cm À1 , see Fig. 1b). Concerning the storage lifetime, a half-life from cis to trans state of 36.3 h in toluene can be extracted from the Eyring plot. The activation enthalpy was calculated as DH ‡ therm ¼ 93.7 kJ mol À1 , together with an activation entropy DS ‡ therm ¼ À31.7 J mol À1 K À1 (S2, ESI †). The quantum yield of the trans to cis photoisomerisation reaction was determined to be 21% at 340 nm, using a literature procedure 44 (S3, ESI †). In addition, it was observed that the AZO1-cis state can photoisomerise back to the trans form. The corresponding quantum yield of the cis to trans photoisomerisation was determined to be 23% at 455 nm, slightly higher than the reverse process. Due to this photo-induced twoway switch effect for AZO1, full conversion via full spectrum solar light can likely not be achieved and a band pass lter was used in the device conversion experiments.
Good cyclability is one of the most important criteria of MOST systems. To investigate the robustness of AZO1, a solution of ca. 10 À5 M in toluene was prepared without degassing. Two controllable LED light sources were turned on and off alternatingly (340 nm with 100 s irradiation time and 455 nm with 300 s irradiation time) to charge and discharge the molecules back and forth. Aer 203 complete cycles (with a total operation time of 22.6 h), no signicant signs of degradation were observed, thus demonstrating high robustness of AZO1, even in the presence of oxygen (see Fig. 1c).
To further demonstrate the functionality of the AZO1 system in a real device, a continuous ow system was built. For this labscale conversion test under an AM1.5 solar simulator, diluted solutions of AZO1 were used to collect data by UV-vis spectroscopy. Two concentrations of AZO1 solutions were pumped individually through a microuidic chip, 20,21 with varying residence time (33.9 mm 3 inner volume with a channel depth of 100 mm, see Fig. 2a). Since AZO1-cis can be back-converted by visible light, an optical UV transmitting band pass lter (<400 nm) was inserted in between the solar simulator and the microuidic chip. The UV-Vis spectra of the AZO1 solution before and aer pumping into the microuidic chip were recorded. Eqn (1) was used to calculate the cis-to-trans conversion percentage (S4, ESI †): where A @350 nm is the actual absorbance at 350 nm; A iso@306 nm and 3 iso@306 nm correspond to the absorbance and absorptivity of the solution at its isosbestic point at 306 nm, respectively; and 3 parent@350 nm and 3 isomer@350 nm are the absorptivity of the AZO1-trans and AZO1-cis isomers at 350 nm, respectively. A maximum conversion of around 80% from the AZO1 trans to cis state was obtained from a solution of 2 Â 10 À4 M. This is likely due to the photo-stationary state of azobenzene trans-to-cis photoisomerization (see Fig. 2b). Concerning the energy storage efficiency, 0.88% of the solar energy could theoretically be stored in a neat sample (S5, ESI †). The highest efficiency that can be expected for a 5 Â 10 À4 M solution is 0.02%, and for a 2 Â 10 À4 M solution is 0.01%. With a band pass lter (SCHOTT, UG11) and a solar simulator, the actual measurements showed that the maximum energy storage efficiency was ca. 0.009% for the 5 Â 10 À4 M solution, and 0.005% for the 2 Â 10 À4 M solution (see Fig. 2c, S6, ESI †). These experimental results were close to reach the theoretical predictions, but are still limited overall by the concentrations used and the photo-stationary state between the two species. Aer solar capture/storage, energy release is the second fundamental process for the MOST concept. To estimate the adiabatic heat release as a function of concentration, a modied equation from a previous study 11 was used (eqn (2)): where c is the concentration of AZO1, M w represents its molecular weight; DH storage corresponds to the DSC measured energy storage capacity of AZO1-cis which equals to 167.5 J g À1 ; r AZO1 is the volumetric mass density of AZO1 supposed to be 1.09 g mL À1 ; C p_AZO1 is the specic heat capacity of AZO1 in J g À1 K À1 , assumed similar to that of unsubstituted azobenzene; 45 and r solvent and C p_solvent address the volumetric mass density in g L À1 and the specic heat capacity in J g À1 K À1 of the solvent, respectively (which are equal to 867 g L À1 and 1.7 J g À1 K À1 ). In this approach, the volume load factor of the solvent and photoisomer was considered, i.e. when the concentration approaches neat conditions, 1 À cM w r AZO1 r solvent C p_solvent approaches zero. This correction is reasonable for all MOST heat release estimations with close to neat conditions. In the case of AZO1, the theoretical maximum temperature difference was calculated as ca. 226 C for a fully charged neat sample (see S7, ESI †). Concerning related catalysts, several candidates, including mineral acids like perchloric acid, Cu(II) salts (CuCl 2 ,Cu(OAc) 2 ), gold nanoparticles which involve a redox mechanism, as well as electrocatalytic methods have been reported to induce the backconversion of azobenzene derivatives. [46][47][48] However, for a closed cycle system which can be operated in devices, a heterogeneous catalyst that can be xed in a reaction centre is required. Based on this fact, a heterogeneous catalyst or an insoluble homogeneous catalyst needs to be developed. For AZO1, two potential catalysts, cobalt(II) phthalocyanine physisorbed on the surface of activated carbon (CoPc@C) and [Cu(CH 3 CN) 4 ]PF 6 fulll the described physical properties and thus, they were tested individually. Both showed a positive effect on reducing the backconversion half-life at room temperature. [Cu(CH 3 CN) 4 ]PF 6 has a very low solubility in toluene, being active for various MOST systems, including norbornadiene/quadricyclane derivatives 11 and dihydroazulene/vinylheptafulvene couples. 20,49 For the AZO1-cis compound, a back-conversion reaction rate of up to 30 s À1 was calculated at room temperature of 25 C (up to 6 Â 10 6 times higher than a reaction rate of 5 Â 10 À6 s À1 without the catalyst at 25 C; see Fig. 3a, S7, ESI †). CoPc@C was produced following a reported procedure; however it had a risk of leaching from its solid support. 11 Therefore, the [Cu(CH 3 CN) 4 ]PF 6 salt was chosen to be incorporated into the catalytic device for further testing. With this result in mind, a small-sized reaction centre was built. 5 mg of Cu(I) salt was inserted into a Teon tube which has a 1 mm inner diameter (see Fig. 3b). 5 Â 10 À4 M 79% AZO1-cis solution from the microuidic chip experiments was then owed through the catalytic bed with a speed of 1 mL h À1 . As result, 48% of AZO1-cis was successfully back-converted to the corresponding trans state. Thus, the described continuous uidic chip experiments can achieve the requirements for the complete application of the MOST concept including photon capture/storage and energy release processes.
To further understand the mechanism of back-conversion with [Cu(CH 3 CN) 4 ]PF 6 , a detailed study within the framework of the density functional theory (DFT) was performed. A simplied AZO1 system was used replacing the ethylhexyl moiety with a methyl group at the PCM-B3LYP-D3BJ/6-31G(d)+SDD(Cu) level of theory (S8, ESI †). Among the different coordination options, the CH 3 CN ligands could be displaced by the new ligand (the cis isomer of the azobenzene moiety, in this case). Four different coordination possibilities were considered to ensure a good modelling of the experimental conditions. In all cases, the formation of the new complex is slightly endergonic (1-5 kcal mol À1 ). As the azobenzene derivative is not symmetric, two sets of coordination alternatives arise. Coordination through the nitrogen atom directly linked to the unsubstituted phenyl ring (path b, S8, ESI † and Fig. 4a) is favored (by ca. 2.3 kcal mol À1 ) as compared to the methoxyphenyl linked nitrogen atom (path a). Isomerization of azobenzenes is very dependent on substitution and reaction conditions. 50 In this case, different pathways were considered and rotation and inversion mechanisms of the isomerization were explicitly evaluated. As in the not catalyzed thermal reaction, 47 the inversion transition states (TS1a-TS1b) were found to be preferred.
A relaxed scan along the rotation coordinate (CNNC dihedral) (see Fig. 4b) reveals a possible S 0 -T 1 -S 0 mechanism accessible by coordination of the nitrogen atom of AZO1-cis to copper, which reveals a MLCT (metal-to-ligand charge transfer). In turn, this implies the participation of a p* orbital of the azobenzene allowing rotation along the N]N with participation of the T 1 state. At relative high torsion angles, the triplet state becomes less energetic and this path yields a barrier of around 20 kcal mol À1 . This alternative mechanism allows for the bypassing of the more energetic inversion TS in the ground state (30 kcal mol À1 ) and it should be available at room temperature.
Thus, copper coordination yields a decrease in the isomerization barrier of ca. 6 kcal mol À1 (computed barrier decreases from 25.4 kcal mol À1 for free azobenzene to 19.4 kcal mol À1 ). This relatively small decrease in the energy barrier is enough to allow the isomerization process to occur in a few minutes instead of some days, thus accelerating the isomerization by 4 orders of magnitude, resulting in good agreement with experiments (energy difference of 8 kcal mol À1 , from 23.5 kcal mol À1 for the free AZO1 to 15.6 kcal mol À1 for the copper activated reaction). Overall, while [Cu(CH 3 CN) 4 ]PF 6 is useful to increase the reaction rate notably, there is still great potential for improvement in the design of the energy release step. This could be done, for instance, favoring the MLCT or stabilizing the MECP between singlet and triplet states to increase the efficiency of the crossing. In both cases, this could lead to a decrease in the energy barrier and a subsequent acceleration of the back-conversion.

Conclusions
An AZO1-MOST charging/discharging cycle has been successfully demonstrated, including the study of the photophysical properties and back-conversion of the photoisomer. The AZO1 system features strong absorption (3 Max@350 nm ¼ 2.6 Â 10 4 M À1 cm À1 ) and long thermal half-life (36.3 h). The optical cyclability test for a diluted solution shows no signicant degradation in the presence of oxygen, allowing the use of AZO1 in toluene in application. For MOST application purposes, the conversion quantum yield at 340 nm was determined to be 21%  which is similar to the quantum yield of back-conversion of 23% at 455 nm. This implies that even diluted samples cannot be fully converted under sunlight. The estimated maximum energy storage efficiency for AZO1 in the neat state was calculated to be 0.88%. To experimentally demonstrate the functionality of this liquid azobenzene for MOST applications, conversion experiments were performed at different concentrations of AZO1 in a continuous microuidic chip system. The measured maximum energy storage efficiency for a 5 Â 10 À4 M solution can reach close to 0.009%, close to reach the theoretical prediction of 0.02% at such a concentration. This, unfortunately, showed a strong photo-stationary effect between AZOcis and AZO-trans, therefore highlights the need to further develop the azobenzene system by altering the optical properties. [51][52][53] Further improvement to increase quantum yields and diminish the absorption of photoisomer, as well as differentiate the spectral difference between cis and trans states, would be certainly required. To estimate the theoretical heat release temperature difference, a corrected formula was introduced to account for the change from the diluted solution to the neat sample. A maximum temperature increase of 226 C can be theoretically achieved by using eqn (2) with a correction term for a neat MOST sample. Furthermore, two new catalysts have been identied to allow the back-conversion of AZO1-cis and a reaction centre based on a Cu(I) salt has been prepared as a proof of concept for the energy release process. With a concentration of 5 Â 10 À4 M and a ow speed of 1 mL h À1 , 48% of the cis state isomer was successfully converted back to the trans state. However, the reaction rate was still too low to be used for macroscopic heat release purposes. Finally, a detailed theoretical study of the mechanism has been proposed to further understand the discharge process using a Cu(I) salt. This could help in the future design of new azobenzene derivatives and the corresponding catalysts. In addition, results shown here could be extended to the control of azobenzene derivatives for different applications.

Conflicts of interest
There are no conicts to declare.