Status and challenges for molecular solar thermal energy storage system based devices

Molecular solar thermal energy storage systems (MOST) oﬀer emission-free energy storage where solar power is stored via valence isomerization in molecular photoswitches. These photoswitchable molecules can later release the stored energy as heat on-demand. Such systems are emerging in recent years as a vibrant research field that is rapidly transitioning from basic research to applications. Since a major part of the attention is focused on molecular design and engineering, MOST-based device development has not been systematically summarized and introduced to a broad audience. This tutorial review will discuss the most commonly used and developed devices from a chemical engineering point of view. It is expected that future developers of MOST technology could be inspired by the existing devices, keeping in mind the summarized essential practical challenges towards large-scale implementations and more innovative applications.


Introduction
In the next 30 years, the world population is expected to increase by 25% to 9.9 billion people. 1 This population increase will come with a continuous increase in energy consumption. Fossil fuels have been the primary energy source over the last decades to cover the energy demand. However, as emissions from the burning of combustible energy sources cause serious environmental effects, such as global warming and air pollution, the latter is also related to health problems, 2 renewable energy sources, such as wind, geothermal and solar power, are advancing due to their potential as green and sustainable alternatives. 3 The sun is one of the most abundant energy sources on Earth. The absorption and scattering of incoming photons bring approximately 3.4 Â 10 6 Exajoules of solar energy to the Earth's surface every year. This is 7000-8000 times the annual world primary energy consumption. 4 Hence, solar energy technologies can potentially meet the growing global energy demands without releasing harmful emissions. However, the use of solar power is still relatively low compared with other energy resources. In 2019, solar power contributed less than 3.6% to the world's total energy production. 3 One of the limiting factors of solar energy is the intermittency of its production and storage. Hence, more significant implementation of solar power necessitates new scalable technologies for solar energy storage. To provide sufficient and reliable energy, new solar energy storage methods are needed.
Several methods for storing solar energy, such as the use of electrochemical batteries, hydrogen energy storage, and carbon dioxide conversion, are being implemented. 5 A relatively unexplored method is the use of photoswitchable molecules, called molecular solar thermal energy storage systems (MOST) or solar thermal fuels (STF), which can directly convert and store solar energy as chemical energy. When energy is needed, the high energy photoisomer can release the stored energy on demand in the form of heat. The molecular design of MOST has been reviewed several times in recent years. [6][7][8] The main purpose of this review is to describe the current status and challenges of MOST-based devices with regard to energy capture, storage, and release based on different molecular systems.
MOST devices can be divided into three different types: (i) charging, (ii) discharging (heat release), and (iii) hybrid (combining MOST with other energy conversion systems). In the following sections, we will briefly describe the functional principles of the MOST system before introducing the most common and recently developed devices. After this, we will introduce the applications of photoswitchable molecules beyond traditional MOST devices. Finally, we will discuss the challenges of MOST devices and briefly compare them with other solar technologies. We hope this review of existing devices will inspire future MOST technology developers to generate more efficient, practical, real-life applications of solar energy.

From MOST concept to practical devices
The MOST concept originated in 1909, when Weigert suggested photodimerizing anthracene molecules to store solar energy. 9 In 1979, Xuan et al. systematically listed the operation principles of the MOST concept and several criteria that a molecule should fulfill. 10 This checklist of molecular properties was refined by Yoshida in 1985, 11 Chernoivanov et al. in 1991, 12 and others. 6,13 Here, we define organic photoswitchable molecules for MOST as compound A (parent, see Fig. 1A) that can undergo a photoninduced chemical reaction to form photoisomer B (photoisomer). B can then be triggered to switch back to A by light, heat, a catalyst, or another method. In MOST systems, this backconversion of B to A is associated with the release of stored energy as heat. As shown in Fig. 1B, the charging (or conversion) and discharging (or back-conversion) reactions have different energy levels and processes. The parent state molecule is first excited by a photon with energy E nm (or hn onset ) and then photoisomerized to the corresponding high energy metastable photoisomer. Here, Gibbs free energy can be simplified as enthalpies since the activation entropies of most of the isomerization reactions are very low. 13 Therefore, the energy difference DH storage between the parent and the photoisomer corresponds

Helen Hölzel
Helen Hölzel is an experimental chemist with experience in synthetic organic chemistry. At present, Helen works as a Post-Doc in Kasper Moth-Poulsen's group, on flow chemistry, the synthesis and photochemistry of molecular switches, and is involved in several projects concerned with the development, optimization, and investigation of the NBD/QC isomerization pair as an efficient MOST system. She studied Molecular Science and obtained her BSc and MSc from the FAU Erlangen-Nuernberg in Germany. She received her doctoral degree with honors (summa cum laude) in natural sciences (chemistry) from the FAU, working on carbon-rich p-system porphyrin hybrids in the group of Prof. Norbert Jux. During her doctoral studies, she also joined the group of Prof. Eiichi Nakamura at the University of Tokyo, Japan, as visiting researcher (JSPS fellowship) where she investigated the behavior of molecular systems under TEM (transmission electron microscopy) conditions.

Tutorial Review
Chem Soc Rev to the energy the couple can store. When the stored energy is required, the photoisomer needs a thermal trigger to overcome the activation energy barrier E a (or DH ‡ therm ). During this charging and discharging process, a part of the energy will be lost due to the thermal relaxation effect (E l ).
Several properties of the MOST systems are strongly correlated, e.g. addition of larger substituents to improve E nm leads to an increase in the molecular weight, thus lowering the energy storage density. That is why there is still room for improvement in molecular design, even though the concept was conceived as early as 1909. 6 Yet, it is still very challenging to chemically design an ultimate candidate that can fulfil all criteria of the MOST concept. For instance, by adding additional push-pull units to red-shift the absorption of a parent state molecule, DH storage can also be lowered. 6 So far, researchers have focused on improving anthracene, azobenzene, norbornadiene, dihydroazulene, and fulvalene-tetracarbonyl-diruthenium systems. 14 In addition to optimizing molecular systems, MOST-based devices are also being developed. Since the operating mechanisms of MOST systems are different from those of traditional energy systems, special device concepts need to be developed. In the following sections, we review the device engineering side of the MOST concept, especially how molecular properties can be addressed and linked to the device, and how they affect the performance of the devices.

MOST charging devices
First, we will consider the theoretical limit of energy storage efficiency of MOST molecules. This efficiency limit implies the maximum efficiency of solar energy storage that is possible for a given MOST molecule. Here, the energy storage efficiency can be calculated as [15][16][17] : N is the number of absorbed photons at an excitation energy E nm per unit time and irradiated area. E a is the thermal energy storage barrier as indicated in Fig. 1B. E l represents the relaxation energy lost during the photoisomerization process, which corresponds to the energy difference between the excited state and the maximum transition state via thermal back-conversion. E AM1.5 corresponds to the total incoming solar power with an air mass (AM) of 1.5. As an initial approach, it is assumed that all incoming photons from the solar spectrum above a molecular S 0 -S 1 excitation threshold are absorbed by the parent molecule. Based on eqn (1), it can be estimated that the solar energy storage efficiency can reach up to 12.4% at an S 0 -S 1 gap of 1.81 eV with a storage time of 24 days at 25 1C. 15 Later, it was shown that, by combining multiple optimized photoswitches with different onsets of absorption, the efficiency can be increased to E21%, where the remaining energy of the solar spectrum is either thermalized during the absorption process or not absorbed by the molecular systems. 16 This multijunction design has been used for both liquid and solid MOST conversion experiments. 16,18 In the experimental situation, additional parameters such as concentration, photoisomerization quantum yield, and absorption of photons by the photoisomer must be considered. To do this, eqn (1) can be reformulated as 19 where T(c, l) is the concentration-and wavelength-dependent transmittance in a specific device with a specific optical path length. f iso is the experimental photoisomerization quantum yield of the parent molecule. This equation only focuses on the optical and energy storage properties of the molecular species. In practice, optical losses, solvent absorption, and thermal back-reaction may lower the efficiency even further. 19 For instance, from eqn (2), the onset of absorption of the molecule can affect T(c, l). This is therefore an important limiting parameter of solar energy storage efficiency. The maximum energy storage efficiency of a norbornadiene derivative NBD1 was estimated to be 0.51% with an onset of absorption at 380 nm. 19 As an example, this efficiency can be improved to 3.8% by applying two NBD units to form NBD2 (see Table 1). 20 Besides, low f iso and/or low DH storage can also significantly affect the efficiency of the system, and these will be described later when we discuss (fulvalene) diruthenium derivatives and azobenzene derivative tests.
The MOST system has a relatively free form-factor in a liquid state, and it can be conveniently transported via pumping. 15 To determine the experimental solar thermal energy storage efficiency for fluidic MOST, the following equation can be used: 23 where : n parent represents the flow speed of the parent molecule in mol s À1 and S is the effective irradiated area of the device in m 2 . a conversion corresponds to the conversion ratio from parent state to photoisomer. This ratio can be calculated using Lambert-Beer's law when the absorptivity of the photoisomer is negligible compared with that of the parent state. If this is not the case, the ratio  can also be calculated with the isosbestic point: where c isomer and c tot are the concentrations of the photoisomer and the whole MOST solution, respectively; A @l is the experimental absorbance at a certain wavelength l; A iso represents the experimental absorbance at the isosbestic point; e iso , e isomer@l , and e parent@l are absorptivities of the specific isosbestic point, photoisomer at l, and parent at l. A major practical challenge of current MOST molecules is that the absorption spectrum of the parent state overlaps with that of the photoisomer and this can significantly affect device performance. From eqn (3), it can be seen that the experimental solar thermal storage efficiency Z MOST can vary by the conversion ratio a conversion . This factor can be interpreted as the difference between the absorption spectrum of the parent molecule and that of its corresponding photoisomer. When the spectra largely overlap, an absorption competition can occur during conversion, increasing the time it takes for the solution to be fully charged. As a result, Z MOST will drop during charging. This intrinsic problem generates an ''inner filter effect''.
Since ideal zero-spectrum overlap molecules have not been identified yet, low concentrations for practical tests are the most convenient way to demonstrate the basic function. The solar collector for charging these systems can be designed in three different ways with increasing light-concentrating capacities (see Fig. 2). 21,30 Firstly, the photoconversion device can be made in a flat panel with meander-shaped channels inside (see Fig. 2a). In laboratory experiments, this only requires a calibrated solar irradiation source with appropriate irradiance (normally with AM 1.5) on top of the panel and different fluidic residence times to accomplish the tests. Unfortunately, comparing the molecular charging performance between studies is not easy since different charging devices and molecular concentrations were used. In 2012, Segalman et al. used the first microfluidic chip device with B50 mm optical depth and a total surface area of 585 mm 2 to convert a soluble derivative of a (fulvalene) diruthenium compound (RuFv1, see Table 1) in toluene. 24 The diruthenium compound absorbed photons at around 450 nm and showed an energy storage density of 95 kJ mol À1 . Unfortunately, the quantum yield was very low (0.2%), so most of the absorbed photons were wasted as molecular relaxation to heat. With a 0.1 M toluene solution, a conversion of 81% and an energy storage efficiency of 0.01% were reported (see Fig. 3).
In 2019, Mikkelsen et al. established a device simulation framework to predict thermal energy storage rates, conversion percentages, and temperature increases for a given MOST system. 31 Their model was based on microscopic parameters including extinction coefficients, photoisomerization quantum yield, and energy storage density. This model has been used to simulate the conversion of a dihydroazulene derivative (DHA1, see Table 1) in a microfluidic chip of given size. The simulated results were confirmed by an experimental DHA1 conversion. 30 This experiment was performed using a fused silica microfluidic chip with a channel depth of 85 mm, an effective exposed volume of 35.72 mL, and an irradiation surface of 5.18 cm 2 . The onset of absorption by DHA1 was 450 nm, with a f iso of 60% in toluene and a calculated energy storage density of 0.14 J g À1 . Using 1 mM and 4 mM concentrations of DHA1 in toluene, different conversion percentages were successfully measured via various fluidic residence times, and a maximum energy storage efficiency of 0.1% was demonstrated with in-line flow cells and a spectrophotometer. 30 Later, Moth-Poulsen et al. designed a similar microfluidic chip solar collector with 100 mm deep channels to test the various conversion conditions of azobenzene derivatives, including AZO1 (see Table 1 and Fig. 4a). 26 AZO1 absorbed photons from 514 nm, which was higher than the absorption of the diruthenium complex. AZO1 also gave a higher photoisomerization quantum yield of 21% and a DH storage of 167.5 J g À1 . However, AZO1 is a positive photochrome and the photoisomer absorption spectrum is more red-shifted than that of its parent state, so it can also be back-converted by visible light (see Fig. 4b).  Consequently, it is difficult to achieve 100% conversion. To address this, a bandpass filter was used to prevent the lightinduced back-conversion from its cis to trans state. Based on these findings, negative photochromism (in which the photoisomer has a more blue-shifted onset of absorption than the parent state) is preferred for MOST applications because of the less inner filter effect. From the experimental results, even with low concentrations of 0.2 mM and 0.5 mM in toluene, the maximum conversion was around 80% in both cases, with an experimental maximal energy storage efficiency of 0.01% (see Fig. 4c and d). Recently, inspired by multijunction solar cells, a liquid-based multijunction MOST device was also experimentally demonstrated and it showed a total energy storage efficiency of 0.02% with a triple microfluidic-chip system. 16 The overall energy storage efficiency of the whole operating device was higher than the efficiency of any of the single-layer devices, suggesting that this may be a way to improve efficiency in future devices.
Similar to traditional reflecting solar concentrators, MOST charging devices can also be made using a parabolic reflecting mirror. 19,22,30 For liquid MOST systems, fluidic tubing can be installed in the focus line of the parabolic collector (see Fig. 2b). In this design, the incoming photons are reflected towards the central area, where MOST liquids are pumped through. 22 The central focusing glass tubing can be heated because of the concentrated photon flux (concentration capacity of 5-30 suns). However, this can be inconvenient for MOST applications, since the photoisomer half-life is a temperature-dependent variable. 6 The hotter the environment is, the shorter the halflife will be. To solve this issue, the central tubing can be changed to a double-layer glass tubing; the MOST solution flows through the outer tubing and cooling water flows through the inner tubing to stabilize the system (see Fig. 5a and b). Studies have been carried out with DHA1 and NBD1 under solar irradiation in Sweden. 19,30 In these studies, the outdoor conversion temperature was cooled to ambient temperature   to maintain half-lives at the same length. With 4 mM DHA1 and 4 mM NBD1 in toluene solution, energy storage efficiencies of 0.02% and 0.03% were achieved, respectively. Later, solar collectors that can concentrate solar energy in a very limited area with a parabolic dish have also been proposed. However, the concentration ratio of these collectors can be 100-10 000 suns, which would damage the MOST molecules. Therefore, cooling would be a major issue for practical MOST applications. So far, this type of solar collector has only been incorporated in theoretical calculations as a part of conceptional devices, which will be discussed later in Section 5. 32,33 In current converting devices, highly concentrated solutions can amplify the inner filter effects, making the conversion process extremely challenging. For instance, converting concentrated parent molecules for further energy release using the two previously mentioned setups can be time-consuming. A possible solution is to use a continuous flow conversion setup with powerful lamps as the source of photons (see Fig. 2c). In 1983, Marangoris et al. established a setup for the continuous conversion of unsubstituted NBD2. 34 In this design, the NBD2 solution was pumped through a glass tube surrounded by a concentric cylindrical UV low-pressure Hg lamp. The sample was continuously charged over cycles and stabilized in a constant temperature bath. Similar to this, Yamakita et al.
(1987) designed a small-scale continuous irradiation system. The system also continuously pumped the solution inside the loop so that the unsubstituted AZO2 (see Table 1) cyclohexane solution was charged over various cycles to enhance conversion to the Z-state. 35 Recent progress in flow chemistry has led to the development of commercially available photoreactors with a highintensity LED light source, which can simplify the conversion process. Moth-Poulsen et al. successfully prepared a 0.8 M NBD4 (see Table 1) toluene solution using a UV-LED irradiation source (365 nm) of 16 W and efficiently reduced the conversion process to 30 minutes (flow rate 0.33 mL min À1 , see Fig. 6). 21 In addition to devices applied for charging purposes, there are also devices for fundamental investigations. For example, Libuda et al. designed a high-intensity UV-photon source operating in an ultra-high vacuum environment (UHV) to study in situ photochemical reactions in organic thin films on a controlled surface (see Fig. 7a). 36 The irradiation source (365 nm) was mounted in a UHV infrared reflection absorption spectroscope, allowing the photochemical process to be tracked in situ on thin films by infrared spectroscopy, inspiring further MOST film development. In this work, unsubstituted NBD3 (see Table 1) was converted with the help of photosensitizer 4,4bis(dimethylamino)benzophenone (Michler's ketone). Herein, the conversion was observed for the first time in thin films (see Fig. 7b). The use of a sensitizer is an alternative way to red-shift the absorption of the MOST system and offers the opportunity to do this without having to engineer the MOST system itself.

MOST heat release devices
Normally, energy can be discharged from MOST systems in four different ways: thermal discharge, catalytic discharge, optical discharge, and electrochemical discharge.
First, one needs to consider the parameters influencing the maximum heat release temperature from a charged MOST candidate. The maximal temperature increase under adiabatic conditions in a neat sample can be calculated by the energy storage enthalpy and specific heat capacity of the photoisomer as shown in eqn (5): However, MOST candidates are generally in a solid state and need a solvent to become liquid for devices. Therefore, the heat capacity of the used solvent has to be taken into account: Here, c is the concentration of the MOST solution and M w is the corresponding molecular weight. C isomer refers to the  specific heat capacity of the material in J g À1 K À1 . r solvent and C solvent correspond to the volumetric mass density in g L À1 and the specific heat capacity in J g À1 K À1 of the solvent, respectively. When the concentration approaches the neat state without solvent, the 1 À c Á M w r AZO1 r solvent Á C p solvent term is zero, and the whole equation can be written as eqn (5). In practical devices, the heat capacity of the full system, as well as various heat losses, also has to be considered when assessing performance.
Of the four-heat extraction routes, a heterogeneous catalyst is the most convenient way to release energy stored in the MOST system. 11 The catalyst needs to be fix-bedded, reusable, and independent of the MOST solution. Therefore, focus has been placed on device development with catalysts. Catalyst development for different MOST systems is crucial. The turnover number (TON) and turnover frequency (TOF) are commonly used to quantify the features of the selected candidate. The higher the TON or TOF of the catalyst, the more efficient the catalysis is. The most convenient way to design a catalyst reactor is to load the catalyst inside a tube to form a fixed bed reactor, and then flow the charged MOST solution through the reactor to release heat.
In 1988, Yoshida et al. demonstrated a catalyst setup. 37 Using unsubstituted NBD3, cobalt(II) deuteroporphyrin and cobalt(II) tetrakis(p-sulfonatophenyl)porphin anchored on alumina beads coated with polyaminesulfone-A (Co(II)DPIX/PAS/Alumina), they observed a maximum heat release temperature of 58.5 1C (see Fig. 8a-c). Unfortunately, unsubstituted NBD cannot operate under solar irradiation since it starts absorbing photons in the UV range at 300 nm. With a substituted NBD5 derivative (see Table 1), a temperature increase of DT = 50.5 1C was recorded in the central part of a tube-shaped fixed bed catalyst device with an inner diameter of 4 cm and a length of 58 cm (see Fig. 8a and d). 38 In 2019, Moth-Poulsen et al. used cobalt phthalocyanine (CoPc) physisorbed on an activated carbon support as the catalyst for converting NBD1. 19 Only 5 mg of CoPc@C was loaded onto Teflon tubing (1 cm long and 1 mm internal diameter). The catalysis reactor was inserted inside a vacuum chamber to prevent heat dissipating to the surrounding area (see Fig. 9a and b). Macroscopic heat release with various concentrations was measured under vacuum conditions, and the results were calculated using a simplified version of eqn (7) (see Fig. 9c). Temperatures up to 63.4 1C were captured with a catalytic TON of 482 and a TOF of 2.0 s À1 (see Fig. 9d). Similar experiments using the same type of catalysis reactor have also been carried out for DHA1, 30 AZO1, 26 and RuFv1, 24 but these experiments did not record significant heat release.
Thermal discharge is the natural way to back-convert the MOST molecule. However, this might not be the most convenient way to release energy on demand. The charged photoisomer always exhibits a specific half-life of energy storage that follows the basic thermodynamic rules; the higher the temperature the molecule sustains, the faster the thermal discharge can proceed (i.e., the shorter half-life the molecule will have). Not many MOST discharging devices have been designed based on thermal discharge methods because a design objective of a MOST system is a long enough half-life (i.e., months or years). In 1983, Marangoris et al. evaluated a large-scale MOST system that was based on unsubstituted quadricyclane (QC, photoisomerized state of NBD3 in Table 1) using a thermal back-conversion energy-releasing method. The setup is shown in Fig. 10. 34 The NBD3 solution was initially stored in a storage tank (T-1), then pumped to a flat plate solar collector (SC-1) to charge up to the QC form. The charged photoisomer solution was then stored in a second storage tank (T-2). To extract energy, the solution was then pumped by pressurized pump P-2 to heat exchanger HE1. This initiated the  thermal back-conversion, allowing water to be boiled in reactor boiler RB-1 over many cycles. Although the system was technically feasible, it was not economically competitive at the time because the photoisomerization quantum yield (90% with sensitizer for unsubstituted NBD3) and the solar energy storage efficiency (0.23% for NBD3) were low. In 1984, the same research group developed a larger setup to investigate the kinetic discharge of QC. 39 The sample was heated constantly at high temperatures (139 1C to 189 1C) and analyzed only periodically using gas chromatography.
Another way to release energy in MOST systems is by optical discharge at a specific wavelength. This method has mainly been used for azobenzene-based compounds in film samples because of the photoreactivity of the Z-form of the system. LEDs or lasers have been used to trigger the release of stored energy. Recently, the azopyrazole derivative AZO3 has been studied. AZO3 (see Table 1 and Fig. 11a) has combined MOST and phase change material (PCM) properties to further store and release heat in the film state (see Fig. 11b). Similar results have been reported by Li et al. 27 and Han et al. [40][41][42] Instead of domestic heating application, de-icing experiments have been successfully demonstrated by using AZO3 (see Fig. 11c). The idea was initially proposed by Grossman et al. in 2013 for smart window applications, and later modeled in detail by Li et al. and different groups. 20,27,[43][44][45] The final energy-releasing route of MOST systems is electrochemical back-conversion. This technique can be an autocatalytical process. Libuda et al. used Pt(111) electrodes to electrochemically trigger the back conversion of NBD2 via the QC+ intermediate to release stored energy. Later, highoriented pyrolytic graphite electrodes were used with a norbornadiene derivative to switch the molecule with more than 99.8% reversibility. Similar studies have also been focused on unsubstituted AZO2 and derivatives. In these studies, the back-conversion of Z to E was explained by the reductive radical cascade reaction pathway. 6 This energy release route has the potential to release the stored energy efficiently as a catalytic method, while at the same time improving our understanding of the MOST molecular energy landscape.

MOST hybrid devices
As previously mentioned, with a multijunction design, an ideal MOST system has a maximum energy storage efficiency of up to 21% (simulated result). 16 Therefore, it is interesting to investigate how the remaining part of the solar spectrum can be utilized. One way is to combine the MOST concept with other technologies to form a hybrid device that can increase solar energy utilization.
To use more photons from the solar spectrum, triplet-triplet annihilation (TTA) photon upconversion technology has been added to a flat-panel device. This idea has been realized with a fulvalene diruthenium derivative. 25 The upconversion system consists of palladium(II) octaethylporphyrin (PdOEP) as a sensitizer and 9,10-diphenylanthracene (DPA) as an annihilator. Two PdOEP compounds absorb two low-energy photons each from the visible area, and quickly relax after photoexcitation to their lowest triplet state with high quantum efficiency. The two triplet state molecules then transfer their energy to the annihilator DPA via Förster resonance energy transfer. After that, DPA emits a single high-energy photon, which is reabsorbed by the photoswitches for the photoisomerization reaction. The detailed process can be expressed as 1 PdOEP + hn lowenergy -3 PdOEP* 3 PdOEP* + 1 DPA -1 PdOEP + 3 DPA* 3 DPA* -1 DPA* + hn highenergy (9) hn highenergy + Parent -Photoisomer In this work, a photon upconversion layer was put on top of the microfluid chip device with RuFv1 solution (see Fig. 12a-c). This finally enhanced the charging of the device by 130% (see Fig. 12d). However, to amplify the observation, a bandpass filter was used to remove incoming UV light in the sensitizer absorption region.
To utilize the full solar spectrum, Lei et al. suggested combining MOST with the thermochemical process. The concept was to concentrate the incoming sunlight with a parabolic solar collector on a fluidic MOST device with NBD6 solution  (see Table 1). 32 In theory, some solar energy can firstly be stored as chemical energy inside the molecules. The transmitted light can then be absorbed by a thermochemical reactor where methanol (CH 3 OH) decomposition produces hydrogen and carbon oxide gas. This hybrid system has an estimated maximum solar to chemical efficiency of 69% (see Fig. 13a). This concept was further developed by adding a photovoltaic system (see Fig. 13b). 33 Unfortunately, this did not improve the total solar utilization efficiency. The system only attained 67% solar utilization efficiency under these conditions, where 7% of the incoming solar energy was stored in MOST and 45% was transformed into the thermal chemical process. The remaining 15% of the photons were converted into electricity. Interestingly, the estimated daily average solar utilization of this device was calculated to be 58% in summer and 45% in winter (in Hebei area, China), suggesting that this type of design can be a promising renewable energy strategy.
Hybrid devices have also been tested experimentally. In 2017, Moth-Poulsen et al. developed a hybrid device that combined solar water heating and MOST. 23 In this study, a fluidic fused silica-based MOST chip was set on top of a water circulating device that was 3D printed using plastic with a quartz cover. A triple-layer insulating mask composed of cardboard, insulating plastic, and aluminum foil was used to prevent unexpected heating from the solar irradiation (AM 1.5). Using NBD7 (see Table 1) with an absorption onset of 456 nm, up to 1.1% of the incoming solar energy was stored as chemical energy without affecting the solar water heating. This led to a combined solar utilization efficiency of up to 80% (see Fig. 14).
In another example, Ghasemi et al. coupled a MOST device with phase change materials for full solar spectrum utilization. 18 NBD1 was coupled with localized phase change materials with carbonized rayon (L-PCM, 54.1% KNO 3 , 20.0% NaNO 3 , 25.9% LiNO 3 ) for potential 24/7 energy delivery (see Fig. 15a). Technically, MOST molecules and phase change materials were isolated by low thermal conductivity in a highly transparent silica aerogel to maintain an appropriate temperature difference (see Fig. 15b). The system further uses a heat transfer fluid (HTF) to harvest the heat. During the daytime, a valve that connects to the MOST layer is closed so that the MOST material

Tutorial Review
Chem Soc Rev could charge under solar irradiation. The valve on the bottom PCM layer is open, allowing the HTF to flow through the L-PCM to transport the released heat to be used for further heating (see Fig. 15c). During the nighttime, the upper valve is open; the charged MOST material releases the stored energy as heat, initiated by the threshold temperature from the L-PCM transferred by the HTF. This device harvested 73% of the 2 kW m À2 solar energy during the daytime. In a large-scale device, the twodimensional side loss can be ignored, and the maximum harvesting efficiency can reach 90% (see Fig. 15d). MOST systems have also been combined with material-based microelectromechanical system thermoelectric generators (MEMS-TEG) to generate power. Moth-Poulsen et al. used an NBD derivative solution (0.78 M NBD4 in toluene, DH isom = 93 kJ mol À1 , 0.37 MJ kg À1 ) and an AZO derivative (AZO3, DH isom = 52 kJ mol À1 , 0.17 MJ kg À1 , DH phase_change = 50 kJ mol À1 , 0.16 MJ kg À1 , see Table 1) as a neat phase-changeable film together with an ultra-thin TEG to generate electric power up to 1.3 W m À3 (see Fig. 16a and b). 21 Using both catalytic (CoPc@C fix-bedded) and optical (LED light irradiation) triggers for the back reaction, this system generated power continuously for almost 30 minutes (see Fig. 16c and d), demonstrating for the first time that MOST technology can transfer solar energy into electricity beyond time and geographical restrictions, and can thus be used as a new type of carbon-based heavy metal free solar battery.   In recent years, more and more studies have combined MOST with other technologies, especially wearable fabrics. For example, in addition to MOST devices, color displays have also been investigated with azobenzene derivatives and phase change materials. Wang et al. presented a series of photochromic azobenzene dopants on PCMs with different molar ratios. The energy density of the system reached 240 J g À1 (Fig. 17a and b) 28 and more than 6 1C of released heat was captured by the thermal camera. Meanwhile, since AZO derivatives change color during charging, a correlation was determined between conversion percentage and color change. This has high potential to be developed into a visual solar thermal energy system ( Fig. 17c and d).
The same research group later designed a wearable MOST fabric that utilized the full solar spectrum, including visible and NIR photons. 29 An AZO derivative, similar to AZO1 in Table 1, with phase change properties was encapsulated in polystyrene. This material effectively prevented leakage and volume change during phase transition of the AZO derivative. The nanocomposites were dispersed with Cs 0.32 WO 3 nanoparticles and applied to the fabric surface. Such Cs 0.32 WO 3 nanoparticles can absorb NIR light, thus transferring heat to AZO units, and then initiate the phase change property. Meanwhile, UV photons from the solar spectrum also contributed to the charging process. For discharge, blue photons from solar irradiation released the stored energy in azobenzene-PCM ( Fig. 18a and b). It was determined that the total energy density for this system was 28.5 J g À1 (101.6 kJ mol À1 ), giving a total solar energy conversion efficiency of 5%, and an experimental heat release of 13 1C. Regarding the robustness of such a system, the charging and discharging process was repeated over 50 cycles, showing high durability of the material. Interestingly, an additional photothermally driven lifter device has also been developed (Fig. 18c). A similar idea using a nanoconfinement approach to improve energy density was demonstrated by Cai et al., they used microphase separation of a block copolymer to selectively disperse phase change materials into nanodomains of a polymer film, forming new MOST composite materials. 46 This promising method can be used to develop wearable warm fabrics. For example, the charged composite can release energy with a 460 nm visible light source, giving a maximum temperature increase of 11 1C. Since this heat increase is less than 50 1C, it will not harm human skin for further wearable application design. Besides, it was suggested that such a composite can also be combined with a polyurethane sponge to actuate the water surface under pneumatic electric field changes.

Summary, challenges, and future work
The estimated maximum MOST energy storage efficiency (20.5%) 16 is certainly better than that of photosynthesis (0.1-0.3%). 47 However, MOST systems must be further developed to meet the conversion efficiency of recent solar cells (up to

Tutorial Review
Chem Soc Rev 47.1% with a six-junction flat-plate terrestrial design under 143 suns). 48 Based on the operation principle of the MOST concept, in which light energy is both converted and stored within the molecule without an external energy storage unit, these future devices will be constructed as a heavy metal-free and highly robust system. Unfortunately, if one considers the publication from Thanh Xuan et al. as the first MOST lab-to-site transfer proposal in 1979, 10 no large-scale device has been implemented within the last 40 years. This may be because this area is still in the fundamental phase of basic molecular motif scanning; the performance of an ideal MOST device will depend on the performance of the selected molecular candidate. Further developments in synthetic, physical, and organic chemistry might uncover more red-shifted MOST molecules that have less spectral overlap with their corresponding photoisomers. Based on the economic evaluation by Marangozls et al. in 1983, cycling capacity is an important parameter to consider when developing MOST systems. 34 So far, the main focus on cyclability is still at the molecular property level, 6 and the only device cycling test was accomplished with NBD4 and a fixed-bed catalyst system. 21 Future small-and large-scale cycling tests are certainly needed. Fortunately, improving technology means the cost of large-scale industrial implementation can largely be balanced by the energy saved. In addition, molecular screening by using machine learning and artificial intelligence (AI) technology 49 will also accelerate the discovery of new candidates, and automation of chemistry systems 50 will increase the potential for scaling up. However, unique and niche, smallscale applications, such as wearable MOST cloth or hybrid devices, may be areas where the MOST systems can be applied in a not too distant future.

Conflicts of interest
There are no conflicts to declare.