## Introduction

A structural phase transition is a fundamental phenomenon that changes the functions of solid-state materials. It is well known that changes in temperature and pressure cause structural phase transitions and greatly change the physical properties of the solid. For instance, ferroelectrics and ferromagnets undergo structural phase transitions at the material-specific Curie temperature, resulting in unique electrical and magnetic behaviors1,2,3,4. Shape memory alloys exhibit the shape recovery effect due to the martensitic transition upon heating5,6. Light is also an external stimulus for structural phase transitions. In photo-induced phase transitions, a new crystal phase appears by light irradiation and changes the electrical and magnetic properties rapidly7,8,9,10. Other external stimuli, such as electromagnetic fields and solvents, also induce structural phase transitions11,12,13. Phase transition mechanisms are widely used in memory, switch, and actuation materials14,15; thus, the discovery of a structural phase transition has potential for both basic science and application fields.

With respect to the function(s) associated with the phase transition, we focus on actuation behavior because mechanically responsive molecular crystals are fascinating candidates for next-generation actuators16,17. Structural phase transitions in molecular crystals lead to actuation such as expansion/contraction18,19,20, bending21, jumping22,23, and locomotion24. In addition to the structural phase transition, photochromic reactions also lead to crystal actuation25,26,27,28,29. Although photomechanical motion is mostly limited to monotonous bending, the combination of structural phase transition and photochromic reaction can induce unique bending behavior30.

This study reports on the photo-triggered phase transition, which differs from previous structural phase transitions. The photo-triggered phase transition was discovered in the crystal of (S)-N-3,5-di-tert-butylsalicylidene-1-(1-naphthyl)ethylamine in the enol form [enol-(S)-1], which undergoes enol–keto photoisomerization (Fig. 1). The enol-(S)-1 crystal exhibits structural phase transitions among trimorphic α-, β-, and γ-phases due to temperature change. Upon ultraviolet (UV) light irradiation, a crystal in the β-phase converts to the γ-phase due to the photo-triggered phase transition. The γ-phase achieved by the photo-triggered phase transition forms a unique molecular conformation that cannot be reached via a thermal phase transition. The manifestation mechanism of the photo-triggered phase transition is considered to originate from strain due to the trans-keto molecules produced by photoisomerization. The photo-triggered phase transition enables thin enol-(S)-1 crystals to bend stepwise via light irradiation; this phenomenon has the potential to extend the functions of photo-responsive solid materials.

## Results

### Thermal phase transition

Differential scanning calorimetry (DSC) measurement of enol-(S)-1 crystals was performed in the temperature range from −100 to 60 °C at a rate of 5 °C min−1 with heating followed by cooling (Fig. 2). On heating, the DSC curve showed two endothermic peaks at −72.5 and 39.6 °C, with enthalpies ΔH of 1.23 and 0.20 kJ mol−1, respectively. On subsequent cooling, the DSC curve showed two exothermic peaks at −81.1 and 31.9 °C, with enthalpies ΔH of −1.03 and −0.24 kJ mol−1, respectively. The thermal hysteresis of each phase transition was 8.6 °C at the low-temperature phase transition α ↔ β, and 7.7 °C at the high-temperature phase transition β ↔ γ. The enthalpy at the phase transition β ↔ γ was approximately one-fifth that of the phase transition α ↔ β, both of which were much smaller than that required for melting at 129.4 °C (Supplementary Fig. 1). These α, β, and γ crystalline phases can be defined as polymorphs of enol-(S)-1 crystal, and the polymorphs appear reversibly due to the low- and high-temperature phase transitions.

X-ray crystallographic analyses of an enol-(S)-1 crystal were performed at various temperatures ranging from −100 to 60 °C, verifying the reversible structural phase transitions among α, β, and γ phases (Fig. 3, Supplementary Tables 1 and 2, and Supplementary Data 1). The crystal structures of α- and β-phases belong to the same space group P1. Due to the low-temperature phase transition α → β, the lengths of a- and c-axes changed by + 1.8 and −1.3% (Fig. 3a, c), respectively, and the α, β, and γ angles decreased by 9°, 5°, and 2.5°, respectively (Fig. 3d–f). In contrast to the α- and β-phases, the crystal structure of the γ-phase belongs to the space group P21. Although the lengths of the a-, b-, and c-axes changed slightly, by + 0.5, −0.05, and + 0.2%, respectively (Fig. 3a–c), the lattice angles showed a more significant change; α and γ angles converted to 90° due to a change in the crystal system from triclinic to monoclinic, and β increased by + 10° during the high-temperature phase transition β → γ (Fig. 3d–f).

To compare the molecular conformation among α-, β-, and γ-phases, we examined the independent molecules in a unit cell of each phase, as shown in Fig. 4a–f. Two independent molecules, drawn in green and yellow, exist in the lattice in the α-phase (Fig. 4a, d); here, these molecules are referred to as molecule 1 (green) and molecule 2 (yellow). Molecules 1 and 2 form slightly different dihedral angles in salicyl and naphthyl planes, of 54.54° and 52.71° at −90 °C, respectively (Fig. 4g). Tert-butyl substituent groups of molecules 1 and 2 in the α-phase are not disordered, forming conformation A (drawn in magenta in Fig. 4a, d, and h).

Molecules 1 and 2 in the β-phase converted to a different conformation from those in the α-phase due to the phase transition α → β (Fig. 4b, e). At −80 °C, the dihedral angles of molecules 1 and 2 were 47.78° and 48.26°, respectively (Fig. 4g), and decreased to 46.40° and 47.58°, respectively, as the temperature increased to 20 °C (Fig. 4g). Another difference from the α-phase is that tert-butyl substituent groups of molecules 1 and 2 in the β-phase become disordered, with two conformations, A (magenta) and B (blue) (Fig. 4b). Conformation B is a rotation of nearly 60° from conformation A. The different orientations of tert-butyl substituent groups in molecules 1 and 2 are clearly visible in the view parallel to the salicyl plane (Fig. 4e). As to the occupancy of the disordered tert-butyl group, conformation A is the main conformation for molecule 1; by contrast, conformation B is the major conformation for molecule 2 in the β-phase (Fig. 4e). The occupancy of conformation A continuously decreases for molecule 1, and increases for molecule 2 with a temperature rise in the β-phase (Fig. 4h).

In the γ-phase, one independent molecule exists in the lattice (Fig. 4c, f). The independent molecule forms a dihedral angle of 46.49° and has conformation A as its major occupancy configuration for the disordered tert-butyl group (Fig. 4g, h). This molecular conformation in the γ-phase is relatively similar to molecule 1 in the β-phase, suggesting that the conformation of molecule 2 in the β-phase changed to the same conformation with molecule 1 by changing the dihedral angle and rotating the tert-butyl group.

Although molecular conformations are clearly different among α,- β-, and γ-phases, the associated molecular arrangements are similar (Fig. 4i and Supplementary Fig. 2). Molecular packings of the (100) face in the α-, β-, and γ-phases show that the phase transitions α ↔ β and β ↔ γ progress with a slight displacement of each molecule along the b axis, inducing changes in the α-angle (Fig. 4i). The molecular packings of the (010) and (001) faces are also similar among the three phases (Supplementary Fig. 2). The similarities in the molecular arrangements among α-, β-, and γ-phases can afford structural phase transitions of α ↔ β and β ↔ γ with small transition enthalpies. Due to the structural phase transitions, a plate-like crystal viewed from the (010) face deforms reversibly without any disintegration upon heating and cooling (Supplementary Fig. 3 and Supplementary Movie 1).

### Photoisomerization

Ultraviolet-visible (UV–Vis) absorption spectra were measured with a transmittance method to reveal the photochromic nature of the enol-(S)-1 crystal (thickness: 20 µm) in the β- and γ-phases. Without UV light irradiation at 20 °C, the spectrum of the crystal in the β-phase showed a large peak at 300 nm, and almost no absorption above 400 nm (Fig. 5a). Upon weak UV light irradiation (365 nm, 5 mW cm−2), a new peak appeared at 450 nm due to photoisomerization from the enol to trans-keto form, with an accompanying crystal color change from yellow to orange (Fig. 5a). The difference spectrum, which was obtained by subtracting the spectrum before UV irradiation from that under UV irradiation, clarified that the absorption increased in the range of 375–550 nm, with a peak at 450 nm (Fig. 5c). The half-life of the thermal back-isomerization from trans-keto to enol form was calculated to be 86.2 s, based on the decay curve of the peak intensity at 450 nm after stopping UV irradiation (Fig. 5d). The back-photoisomerization half-life from the trans-keto to enol form by visible light (488 nm, 1.4 mW cm−2) was calculated to be 9.9 s using a diffuse reflectance spectrometer (Supplementary Fig. 4). The spectral behavior indicates that back-photoisomerization proceeds much faster than that of thermal back-isomerization.

At 50 °C, the spectrum of the crystal in the γ-phase under UV irradiation did not change significantly from that before UV irradiation (Fig. 5b). Although the difference spectrum failed to provide any discernible new peak in the range of 375–550 nm, the time dependence at 450 nm indicated that the absorption intensity upon UV irradiation increased slightly from that before UV irradiation, suggesting photoisomerization with considerably lower conversion in the γ-phase than that in the β-phase (Fig. 5c, d).

### Photo-triggered phase transition

Surprisingly, when the unit cell parameters of an enol-(S)-1 crystal (0.4 × 0.4 × 0.08 mm3) were measured during UV irradiation (365 nm, 60 mW cm−2), the unit cell of the β-phase converted to that of the γ-phase (Fig. 6). Here, the change in the α-angle is an indicator of the photo-triggered phase transition, because the α-angle changes the most during the α → β transition, reaching 90° with the β → γ transition. The α-angle in the β-phase changed to 90° under UV irradiation at various temperatures of −50, −20, 0, and 20 °C (Fig. 6a). In contrast to the change that occurred in the β-phase, the α-angle measured at −120 °C (α-phase) and 60 °C (γ-phase) did not change under UV irradiation (Fig. 6a). The directions of the phase transitions are schematically summarized (Fig. 6b). The β- and γ-angles also coincided with those corresponding to the γ-phase, accompanied by a slight length change in the a-, b-, and c-axes (Fig. 6c). These results suggest that UV light irradiation triggers the β → γ transition, but not the α → β phase transition.

Interestingly, when a thick enol-(S)-1 crystal (2010 × 1235 × 737 µm3) was irradiated by UV light (365 nm, 80 mW cm−2) at room temperature (β-phase), we could clearly see the phase transition β ↔ γ under a microscope, as follows. The crystal started to transform from the irradiated (001) surface upon UV irradiation, and then the phase boundary propagated in the opposite direction (Fig. 7a and Supplementary Movie 2). The angle between the (001) and (012) faces at the upper right edge changed from 47° to 45° (Fig. 7a, c). The transformation continued to propagate to a lower direction within 1 s, and the angle between (0$$\bar 1\bar 1$$) and (00$$\bar 1$$) at the lower left edge changed from 67° to 63° (Fig. 7a, c). The deformed shape after UV irradiation corresponded to the crystal shape of the γ-phase (Fig. 7c). The propagation speed of the phase boundary was calculated as 920 µm s−1 at 80 mW cm−2 UV light intensity. The propagation speed slowed with a reduction in the light intensity, and the deformation was incomplete for the weaker light intensity of 20 mW cm−2 (Supplementary Fig. 5). These results indicate that there is a threshold UV light intensity to realize the photo-triggered phase transition.

After stopping UV irradiation, the crystal maintained its deformed shape for 50 s. Then, the phase boundary appeared, with the return of the angle at the lower left edge from 63° to 67° (50 s, Fig. 7b). The phase boundary gradually progressed to the upper direction (70 s, Fig. 7b), and then the entire crystal returned to its initial shape in the β-phase (110 s, Fig. 7b) The propagation speed of the reverse deformation is roughly consistent with the speed of thermal back-isomerization from the trans-keto to enol form (Fig. 5d).

To clarify the photothermal effect by UV light, the surface temperature of an enol-(S)-1 crystal was monitored with an infrared (IR) thermography camera (Fig. 7d, e). When the crystal was irradiated with UV light (365 nm, 80 mW cm−2), the irradiated surface temperature rose from 22.6 to 26.8 °C in 0.2 s, at which the photo-triggered transformation already started (Fig. 7a, e). Then, the surface temperature continued to increase to 37.5 °C in 1 s, at which the transformation completed (Fig. 7a, d, e). Thus, this temperature measurement revealed that the photo-triggered transformation progressed before reaching the thermal phase transition temperature.

After stopping the UV irradiation, the surface temperature decreased to the initial temperature in 5 s (Fig. 7e). The time scale of the temperature decrease was much faster than the crystal shape recovery (Fig. 7b). These results suggest that the photo-triggered phase transition β → γ is triggered by enol–keto photoisomerization, and the reverse transition γ → β occurs due to back-isomerization.

When the photoirradiation continued for 20 s, the surface temperature increased by nearly 30 °C in 5 s, and then kept a steady state during 5–20 s (Supplementary Fig. 6). This result indicates that a crystal temperature will not increase above the phase transition temperature during low-temperature X-ray measurements.

### Crystal structure change upon UV irradiation

To elucidate the structural difference of γ-phase formed by the photo-triggered phase transition and the thermal phase transition, the crystal structures were determined using the same sample shown in Supplementary Fig. 7 before and under UV irradiation at −50 and 50 °C (Fig. 8, Supplementary Tables 3 and 4, and Supplementary Data 2). The crystal structures of the β- and γ-phases without UV irradiation are described in detail in a previous section. Under UV light irradiation at −50 °C, the crystal structure of the β-phase changed to that of the γ-phase; the trans-keto molecule was not found, likely due to insufficient quantities for structural determination. However, the enol-(S)-1 molecule in the γ-phase achieved by a photo-triggered phase transition has a slightly different conformation from that of the γ-phase achieved by simple heating. The molecule under UV irradiation at −50 °C forms a dihedral angle of 47.4°, and the occupancy of the disordered tert-butyl group is A:B = 90:10 (Fig. 8b). By contrast, the molecule without UV light at 50 °C forms a dihedral angle of 46.6°, and the occupancy of the disordered tert-butyl group is A:B = 71:29 (Fig. 8c). The conformational difference reflects on the temperature difference, as the dihedral angle decreases to 46.6° and the occupancy of the disorder approaches A:B = 71:29, depending on the temperature rise (Fig. 4g, h). The molecular conformation under UV light at 50 °C did not change from that before UV irradiation (Fig. 8c, d). Thus, the γ-phase achieved by the photo-triggered phase transition is a unique state with respect to molecular conformation, which cannot be achieved by thermal phase transition. To our knowledge, this is the first evidence of this kind of photo-triggered phase transition.

Subsequently, the time dependence of unit cell changes after stopping the UV light irradiation was measured at −50 °C (Supplementary Figs. 8 and 9). For instance, the α-angle, which becomes 90° due to the photo-triggered phase transition β → γ, maintained 90° at least 80 min after stopping the irradiation (Supplementary Fig. 8). By contrast, when the crystal was irradiated by visible light (488 nm, 10 mW cm−2) just after stopping the UV light irradiation, the α-angle returned to 85° in the β-phase within 20 min (Supplementary Fig. 9). Faster return of the unit cell under visible light corresponds to back-photoisomerization, which is much faster than thermal back-isomerization (Supplementary Fig. 4). These lattice change behaviors, with and without visible light, verify that the phase transition γ → β is triggered by back-isomerization from the trans-keto to enol form.

### Stepwise bending

From the perspective of actuation behavior, the photo-triggered phase transition should lead to stepwise bending (Fig. 9 and Supplementary Movie 3). When the (001) face of a thin plate-like enol-(S)-1 crystal (4072 × 940 × 36 μm3) in the β-phase was irradiated from the left with UV light (365 nm, 60 mW cm−2) at room temperature, the crystal bent towards the light source with a twisting motion in 0.2 s, accompanied by a color change from yellow to orange (Fig. 9a, b). In the next moment, the bent crystal suddenly released its twisting motion (Fig. 9c). Then, the crystal continued to bend towards the light source without twisting under prolonged UV irradiation (Fig. 9d). After stopping light irradiation, the bending gradually returned without twist in 45 s (Fig. 9e). The bent crystal started to twist slowly, and then exhibited its most twisted shape at 70 s (Fig. 9f). The bending with a twist gradually relaxed, and then the crystal returned to its initial straight shape and its color returned to yellow (Fig. 9g).

The bending behavior during the process described above can be divided into three steps: (1) bending with twisting due to enol–keto photoisomerization at the β-phase, (2) the disappearance of the twist due to the photo-triggered phase transition β → γ, and (3) bending without twisting due to enol–keto photoisomerization in the γ-phase. The motion of each step was confirmed by applying weak UV light and IR light heating separately (Supplementary Fig. 10). Thermal relaxation of the bending behavior can be explained by the reverse process. Bending relaxation without twist occurs due to thermal back-isomerization in the γ-phase. Then, twisting motion gradually appears due to the progression of the phase transition γ → β triggered by thermal back-isomerization. The bending with a twist returns to the initial unbent shape due to thermal back-isomerization in the β-phase.

The stepwise bending was repeatable at least 40 cycles without disintegration, but in some cases, crystals cracked along the width direction of a axis (Supplementary Fig. 11). The crack should be generated by the strain of surface contraction along the length direction of b axis, which is consistent to the bending towards the light source.

## Discussion

To consider the structural changes due to photoisomerization, the structures of enol, trans-keto, and cis-keto forms were optimized by density functional theory calculations at the theory level of B3LYP/6-31G (Fig. 10a)31. The optimized structure of enol-(S)-1 with a dihedral angle of 56° was consistent with X-ray crystallographic results for the enol-(S)-1 crystal; cis-keto-(S)-1 was similar to enol-(S)-1. By contrast, the structure of trans-keto-(S)-1 differed significantly from enol-(S)-1, forming a dihedral angle of 85°. This structural difference between enol and trans-keto forms suggests that photoisomerization from the enol to trans-keto form undergoes by pedal motion in a limited space32 and generates strain in the crystal (Fig. 10b).

Photoisomerization should lead to a photo-triggered phase transition (Fig. 10c–e). The enol-(S)-1 crystal before photoirradiation is in the β-phase (Fig. 10c). Upon UV irradiation, trans-keto molecules are produced due to photoisomerization at the irradiated surface, generating strain in the crystal (Fig. 10d). The strain due to trans-keto molecules induces conformational change of the surrounding molecules. The conformational change further influences the adjacent surrounding molecules, similar to a domino effect, triggering crystal structure change from the β- to γ-phase (Fig. 10e). When trans-keto molecules are produced at a certain low ratio, the whole crystal completes the phase transition into the γ-phase. When the light irradiation stops, thermal back-isomerization from the trans-keto to the enol form occurs in the crystal, triggering the reverse phase transition γ → β (Fig. 10d).

This transition mechanism indicates that the γ-phase reached by the photo-triggered phase transition requires trans-keto molecules to maintain the crystal phase. In fact, the phase transition γ → β under visible light irradiation proceeds faster than that without visible light, due to the half-life of back-photoisomerization from the trans-keto to the enol form, which is much shorter than that of thermal back-isomerization.

As such, trans-keto molecules are the source of strain in a crystal and the conversion ratio is crucial to analyze quantitatively the photo-triggered phase transition. However, it is impossible to obtain the conversion ratio in this case, because the pure trans-keto molecule cannot be isolated due to spontaneous thermal back-isomerization. As an estimation, the conversion ratio might be at most 5% or so, based on the insufficient amount for X-ray structure determination. In addition, the generated strain from trans-keto molecules is difficult to calculate due to bending with a twist. As our future work, experimental measurement of strain–stress relationship produced by trans-keto molecules is required for quantitative analysis of the photo-triggered phase transition.

The manifestation of the photo-triggered phase transition should also be correlated with enthalpy, which is the energy barrier required for the phase transition. The enthalpy at the phase transition β → γ (0.2 kJ mol−1) is one-fifth smaller than that at the phase transition α → β (1 kJ mol−1). The difference in these transition enthalpies supports that enol–keto photoisomerization triggers the phase transition β → γ, but does not trigger the phase transition α → β due to a larger enthalpy. The enthalpy at the phase transition β → γ is also smaller than other structural phase transitions of molecular crystals18,21,24,30,33,34,35.

Additionally, enthalpy at the transition β → γ is 10–200 times larger than the energy densities of superelastic and ferroelastic molecular crystals, which can deform with the creation of a new crystalline phase or twin domains by manually applying shear stress36,37,38,39. This suggests that it will be difficult to create the γ-phase by shear stress in the enol-(S)-1 crystal at the β-phase. In fact, we did not see any deformation of the enol-(S)-1 crystal when manually applying stress under a microscope.

Finally, we would like to discuss the difference between the photo-triggered phase transition described in this paper and the photo-induced phase transition that appears in the literatures7,8,9,10. The crystal phase due to the photo-induced phase transition appears only by light irradiation, which changes the electric and/or magnetic properties of the crystals within femto- or picoseconds. Regarding the photo-triggered phase transition, the crystal phase triggered by light is identical to that via a thermal phase transition, induced by heating, but unique with respect to its molecular conformation. Thus, the photo-triggered phase transition may lead to a new strategy to broaden the applicability of photo-responsive solids.

## Methods

### Material preparation

An enol-(S)-1 compound was synthesized using microwaves (Monowave 300, Anton Paar, Graz, Austria), according to the literature40. Plate-like single crystals of enol-(S)-1 were obtained by evaporation of 2-propanol or acetonitrile solution at ambient temperature.

### Thermal analysis

DSC measurements were conducted with a differential scanning calorimeter (Thermo plus EVO2 DSCvesta, Rigaku) at a speed of 5 °C min−1 in the temperature range from −100 to 60 °C. Another calorimeter (DSC 8500, Perkin Elmer, Waltham, MA, USA) was also employed for measurements in the range of 30–160 °C. The molar transition enthalpy ΔH was calculated based on the peak area under the DSC curve from the baseline.

### UV–Vis absorption measurement

Absorption spectra of an enol-(S)-1 crystal were measured using a transmittance method41. A thin plate-like crystal prepared by solvent evaporation was fixed on a Cu plate with a pinhole of diameter of ~0.5 mm. Transmitted light intensity through the sample was measured over the range of 250–680 nm at 2-nm resolution. Transmitted light intensity through the Cu plate without a sample was also measured in the same wavelength range for blanks. UV–Vis absorption spectra were calculated according to the Beer–Lambert–Bouguer law, and then divided by the crystal thickness to obtain the absorption coefficient for standardization. Sample temperature was regulated by a Peltier thermostat. A diffuse reflectance spectrometer (Lambda 650, Perkin Elmer) was also employed to measure the spectral behavior upon visible light irradiation (λ = 488 nm, FOLS-2, Sawaki Kobo).

### X-ray crystallographic analysis

Single-crystal X-ray diffraction data of enol-(S)-1 without UV light at various temperatures were collected using a R-AXIS RAPID diffractometer (Rigaku) equipped with monochromatic Mo-Kα radiation (λ = 0.71075 Å) at 50 kV and 40 mA. The temperature of the sample was regulated using a N2 gas flow cryostat and calibrated with a thermocouple. Unit cell parameters before, under, and after UV light irradiation were also measured with this diffractometer. The light irradiation was performed with a UV-LED lamp (λ = 365 nm, UV-400, Keyence, Osaka, Japan) and LD light source (λ = 488 nm, FOLS-2, Craft Center SAWAKI Inc.). Crystal structures before and under UV light irradiation at −50 and 50 °C were determined using an XtaLAB Synergy diffractometer (Rigaku) equipped with monochromatic Cu-Kα radiation (λ = 1.54187 Å). The crystal structures were solved using a direct method with SHELXS201342 or SHELXD201342, and then refined on F2 using the full-matrix least-squares method of SHELXL42. Calculations were performed using the Rigaku crystal structure software package43 and a graphical interface, ShelXle44.

### Observation of mechanical behavior of crystals

Shape change upon heating and cooling was observed using a temperature control stage (Japan High Tech) and a digital high-speed microscope (VHX-500, Keyence). Shape change and bending behavior of enol-(S)-1 crystals upon UV light irradiation (λ = 365 nm, UV-400, Keyence) were observed at room temperature using the digital high-speed microscope. Surface temperature change upon UV light irradiation was measured with an IR thermography camera (FSV-2000, Apiste) in a super dry room. The substrate temperature of the glass plate was regulated with a Peltier thermocontroller (VPE-35, VICS).