Crystalline-Amorphous-Crystalline Transformation in a Highly Brilliant Luminescent System with Trigonal-Planar Gold(I) Centers

Photoluminescent compounds showing emission color changes in response to external stimuli have received considerable attention because of their wide range of applications. Here, we report the unique photoluminescence behavior of a digold(I) coordination system with trigonal-planar AuI centers, [Au2(dppm)3]2+ (dppm = bis(diphenylphosphino)methane). This system shows an extremely intense phosphorescence, with a quantum yield of >95% in the solid state. Both the emission color and thermal stability vary due to changing counter ions (Cl− vs. OTf−). Of particular note is the thermal crystalline-amorphous-crystalline transformation for the chloride salt, which is accompanied by drastic emission color changes. Single-crystal and powder X-ray diffractions demonstrate that the two-step transformation is induced by the loss of water molecules of crystallization with the subsequent removal of a dppm ligand to form [Au2(dppm)2]2+, which is mechanically reverted back to [Au2(dppm)3]2+.


Synthesis and structural characterization of [1]Cl 2 •8.5H 2 O and [1](OTf) 2 •H 2 O. Treatment of
[Au 2 Cl 2 (dppm)] 19 with dppm at a 1:2 ratio in MeOH/H 2 O produced a clear yellow solution, from which the chloride salt of [1] 2+ ([1]Cl 2 ·8.5H 2 O) was isolated as pale yellow block crystals at a high yield. The OTf − salt of [1] 2+ ( [1](OTf) 2 ·H 2 O) was also obtained as pale yellow platelet crystals by adding NaOTf to the yellow reaction solution. The elemental and thermogravimetric (TG) analytical data implied that freshly prepared crystals of the chloride and OTf − salts contained 8.5 water molecules and one water molecule per one complex cation, respectively. The presence of water molecules in each compound was confirmed by IR spectroscopy, showing a broad band at approximately 3400 cm −1 (Supplementary Fig. 1) 20 .
Single-crystal X-ray crystallography indicated that [1]Cl 2 ·8.5H 2 O is crystallized in a cubic space group Pa-3, consisting of one third of the complex cation of [1] 2+ , two thirds of chloride anions, and two and five sixths of water molecules of crystallization in the asymmetric unit. The entire complex cation contains two Au I ions that are triply bridged by three dppm ligands, forming a digold(I) structure in [Au 2 (dppm) 3 ] 2+ (Fig. 2a). Each Au I ion, which lies on a crystallographic C 3 axis, is coordinated by three P atoms from three different dppm ligands in an ideal trigonal-planar geometry (av. Au-P = 2.41 Å and P-Au-P = 120°). This triply bridged structure in [1] 2+ is reminiscent of that found in the corresponding disilver(I) complex, [Ag 2 (dppm) 3 ] 2+ 21 . The intramolecular separation between two Au I centers in [1]Cl 2 ·8.5H 2 O is 3.03 Å, suggestive of the presence of an aurophilic interaction. In crystal, there exist intermolecular CH···π interactions between complex cations (C···Ph = 3.07-3.50 Å).  Moreover, two Cl − counter anions and water molecules of crystallization are located around each complex cation, forming CH···Cl (C···Cl = 3.56-3.81 Å) and OH···Cl (O···Cl = 2.91-2.96 Å) interactions.
Single-crystal X-ray analysis indicated that [1](OTf) 2 ·H 2 O is crystallized in a monoclinic space group P2 1 /n, consisting of two complex cations of [1] 2+ (av. Au-P = 2.39 Å and av. P-Au-P = 120°), four OTf − anions, and two water molecules of crystallization in the asymmetric unit. The digold(I) structure of each complex cation in this compound is nearly the same as that found in [1]Cl 2 ·8.5H 2 O (Fig. 2b). The intramolecular separations between two Au I centers in the two complex cations are 2.97 and 2.99 Å, which are slightly shorter than the separation in [ Fig. 2), indicative of the phosphorescent character of emission for both compounds. Time-dependent density functional theory (TD-DFT) calculations of [1]Cl 2 ·H 2 O showed intense absorption at 325.6 and 324.8 nm, which involved one-electron transitions from HOMO-11 to LUMO and HOMO-12 to LUMO, respectively. These MOs possessed a large contribution from the Au···Au core ( Supplementary Figs 3 and 4). Moreover, the DFT calculations confirmed that the luminescence of this compound has a phosphorescent character due to an intersystem crossing from the lowest singlet excited (S 1 ) state to the lowest triplet excited (T) state; the calculated emission wavelength (511 nm) of the luminescence was highly similar to the experimental one. Notably, the internal luminescence quantum yields (Φ ) for both salts were estimated to be more than 95% based on the absolute method using an integrating sphere. This value is considerably greater than that for the corresponding doubly bridged digold(I) complex, [Au 2 Cl 2 (dppm) 2 ] 12 ; the quantum yield of this complex was estimated to be 69% (λ em = 480 nm) under the same conditions. The emission band of [1]Cl 2 ·8.5H 2 O and [1](OTf) 2 ·H 2 O shifted slightly (by 10-15 nm) to longer wavelength when the temperature was lowered to 77 K, indicative of a stronger Au···Au interaction in the excitation state at lower temperature ( Supplementary Fig. 5). In addition, the emission lifetime (ca. 5 μ s) observed at 77 K was nearly the same as that (ca. 4 μ s) observed at room temperature ( Supplementary Fig. 2 Supplementary Fig. 6), presumably because of the equilibrium between the triply bridged structure in [1]  indicated that a gradual weight loss of 8.0%, which corresponds to the loss of 8.5 water molecules, occurred until 373 K to give a dehydrated form, followed by its decomposition at 573 K ( Supplementary Fig. 7). Although no weight loss was observed between 373 and 473 K, the differential scanning calorimetry (DSC) showed an exothermal peak at 399 K, indicative of the occurrence of an endothermal reaction at this temperature. In parallel with the dehydration reaction and subsequent endothermal reactions, dramatic changes in the original emission color were recognized; the emission color changed from green to yellow when a solid sample of [1]Cl 2 ·8.5H 2 O was heated at 373 K, and further heating to 399 K caused the color change to blue (Fig. 4a). In the emission spectra, the sample heated at 373 K showed an emission band at 596 nm with a quantum yield of 52%, whereas the sample heated at 399 K showed an emission band at 470 nm with a quantum yield of 55% (Fig. 4b). We carried out a solid-state 31 P NMR spectroscopy and powder X-ray diffraction (PXRD) study to characterize the two species formed by heating at 373 and 399 K. The 31 P NMR spectrum of an original solid sample of [1]Cl 2 ·8.5H 2 O showed a signal centered at δ 18 ppm ( Supplementary Fig. 8), which was assigned to P donors bound to Au I centers. The same NMR spectral feature was observed for the sample heated at 373 K, indicative of the retention of the digold(I) structure in [1] 2+ by dehydration. In contrast, the sample heated at 399 K exhibited a new signal at δ − 44 ppm. Because the signal at δ − 44 ppm corresponds well with that observed for a solid sample of dppm, the heating of [1]Cl 2 at 399 K causes the dissociation of a part of dppm ligands in [1] 2+ . In the PXRD, the diffraction pattern of a solid sample of [1]Cl 2 ·8.5H 2 O was consistent with that simulated from its single-crystal X-ray data. In contrast, no notable diffractions were observed for a sample heated at 373 K. This is indicative of the collapse of the non-covalent intermolecular interactions due to the removal of water molecules of crystallization, thus converting them to an amorphous solid. Remarkably, the PXRD for a sample heated until 399 K exhibited sharp diffractions that were distinct from those for [1]Cl 2 ·8.5H 2 O. The structure of this new crystalline phase was successfully determined by the high-resolution powder X-ray diffraction using synchrotron radiation (λ = 1.30 Å) (Fig. 2c, Supplementary Fig. 9), which revealed the presence of a complex cation of [2] 2+ and two chloride ions in the asymmetric unit. This complex cation contains two linear Au I ions that are bridged by two dppm ligands, and each Au I center has a diagonal P-Au I -P geometry (av. Au-P = 2.34 Å and P-Au-P = 159°), with an Au···Au separation of 2.97 Å. In this structure, Cl − ions are not involved in the coordination (Au···Cl > 3.6 Å) but act as a counter anion with the formation of CH···Cl interactions with neighboring complex cations. This structure is in sharp contrast to the previously reported structure in [Au 2 (dppm) 2 Cl 2 ]•acetone 23 , in which each Cl − ion weakly coordinates to an Au I center (av. Au-Cl = 2.77 Å) to form a T-shaped coordination geometry ( Supplementary Fig. 10). Moreover, complex cations are connected to each other through multiple CH···π or π ···π interactions, forming a closely packed lattice structure. Whereas non-coordinating dppm molecules are not accommodated in this crystal lattice, the elemental analytical data and the solid-state 31 P NMR spectrum implied that the dppm molecules were not sublimated to air but were still contaminated as an amorphous solid in the sample heated at 399 K.
When a blue-emissive crystalline powder of [2]Cl 2 was manually ground in an agate mortar, a yellow-emissive amorphous powder of [1]Cl 2 was produced. Moreover, the continuous grinding of the powder of [1]Cl 2 after adding water produced a green-emissive crystalline powder of [1]Cl 2 ·8.5H 2 O. The assignment of these powders was made using their emission spectra and PXRD profiles ( Supplementary Figs 11 and 12), thus indicating the reverse conversion from [2]Cl 2 to [1]Cl 2 and then to [1]Cl 2 ·8.5H 2 O. Here, no significant change in the emission color was observed for [1](OTf) 2 ·H 2 O, even when its sample was heated to 473 K. Furthermore, the emission spectrum and PXRD pattern of a sample heated at 473 K were essentially identical to those of the original sample ( Supplementary Figs 13 and 14), illustrating that its thermal stability is considerably higher than that of the OTf − salt of [1] 2+ .

Discussion
We succeeded in the crystallization and structural characterization of [ [1](OTf) 2 ·H 2 O was found to have an excellent quantum yield of more than 95%. This high quantum yield appears to be due to the molecular rigidity around two Au I ions tightly bridged by three dppm ligands, together with the multiple intermolecular interactions in crystal, which completely prevented a nonradiative deactivation. The molecular rigidity was supported by the DFT calculations, which revealed a quite small structural difference between the ground singlet (S 0 ) and triplet excited (T)  (Supplementary Fig. 15). More remarkably, [1]Cl 2 ·8.5H 2 O showed a two-step thermal transformation from its green-emissive crystalline phase to the blue-emissive crystalline phase ( [2]Cl 2 ) via the yellow-emissive amorphous phase ([1]Cl 2 ), which was induced by the loss of water molecules of crystallization and the subsequent dissociation of a dppm ligand. The reverse conversion from [2] Cl 2 to [1]Cl 2 ·8.5H 2 O by mechanical grinding in the presence of water via [1]Cl 2 was also recognized. To our knowledge, this is the first phosphorescent system with a high quantum yield of > 95% that illustrates a reversible, crystalline-amorphous-crystalline thermal transformation accompanied by drastic emission color changes. Such a thermal transformation was not observed for [1](OTf) 2 ·H 2 O, which possesses a rigid crystalline framework, sustained by multiple cation-cation and cation-anion interactions without the mediation of water molecules. Thus, it is reasonable to consider that the water-molecule-mediated crystal structure in [1]Cl 2 ·8.5H 2 O, which was converted to the loosely packed amorphous structure in [1]Cl 2 via the removal of water molecules, together with the formation of the closely packed, non-hydrated crystal structure in [2]Cl 2 via the removal of a dppm ligand, is responsible for this unique thermal transformation. Finally, the present study demonstrated that trigonal-planar Au I species are highly available for the future design and creation of functional luminescent materials.

Methods
Materials and equipment. All reagents and solvents used in synthetic studies were commercially available and used as supplied without further purification.
The IR spectra were recorded on a JASCO FT/IR-4100 infrared spectrophotometer using KBr disks at room temperature. The elemental analysis (C, H) was performed at Osaka University using YANACO CHN coda MT-5 or MT-6. The 1 H and 31 P NMR spectra in solution were measured using a JEOL EX-500 spectrometer at the probe temperature, using tetramethylsilane (TMS, δ 0.0 ppm) as the internal standard for 1 H NMR measurements and triphenyl phosphate (δ −17.6 ppm) as the external standard for 31 P NMR measurements. The 1 H and 31 P NMR spectra are illustrated in Supplementary Figs 16 and 17, respectively. The magic angle spinning (MAS) 31 P NMR spectra in the solid state were measured using a Chemagnetics CMX300 spectrometer at the probe temperature and spun at 7 kHz for [1]Cl 2 •8.5H 2 O and at 6 kHz for [1]Cl 2 and [2]Cl 2 , using triphenyl phosphate as the external standard. The TG and DTA measurements were measured using a SHIMADZU DTG-60 analyzer. The PXRD patterns were recorded using a BRUKER D2 PHASER at room temperature. High-quality PXRD patterns for structural determination were recorded at room temperature in transmission mode using a diffractometer equipped with a blue imaging plate detector at the SPring-8 BL19B2 beamline. The crystals were placed in 0.3 mm glass capillary tubes. The samples were rotated during the measurements. The diffraction patterns were collected using a large Debye− Scherrer camera. The powder simulation patterns were generated from the single-crystal X-ray structures using Mercury 3.0.
Preparation During the second heating, the color of the powder changed from yellow to white. The samples thus obtained were used for the emission and powder X-ray diffraction measurements at ambient temperature.
Luminescence measurements. The luminescence spectra were recorded by a JASCO FP-8500 spectrometer at room temperature in the solid state or in solution, using a Xe lamp as the light source. The internal emission quantum yields (Φ ) were obtained via the absolute measuring method using an integrating sphere unit (JASCO ILFC-847), the internal surface of which was coated with highly reflective Spectralon. The ESC-842 Calibrated light source (WI) and the ESC-843 Calibrated light source (D2) were used to calibrate the emission intensities to measure the absolute quantum yields. The accuracy of this instrument was confirmed using a Rhodamine B ethylene glycol solution. The measurement was performed according to the following protocol. First, nothing was set on the sample cell holder in the integrating sphere, and then, the spectrum of the incident light was measured. The observed peak area was defined as the area from incident light, A 0 (equivalent number of photons in the incident light). Second, the sample on the sample holder was set in the integrating sphere, and the emission spectra of the sample were measured. The obtained excitation wavelength peak area was defined as the area scattered from the sample, A 1 (equivalent number of photons that were not absorbed), and peak area in the emission wavelength range was defined as the area emitted from the sample, A 2 . Finally, the internal emission quantum Scientific RepoRts | 6:26002 | DOI: 10.1038/srep26002 yields (Φ ) were calculated using the following equation: Φ = A 2 /(A 0 -A 1 ). We have measured Φ values for more than 10 samples of [1]Cl 2 ·8.5H 2 O or [1](OTf) 2 ·H 2 O. As a result, the observed Φ values varied in the range of 99-104% for [1]Cl 2 ·8.5H 2 O and 99-101% for [1](OTf) 2 ·H 2 O. The fluctuations in the Φ values (maximum of 5%) should be regarded as the random error of measurements. In addition, we have measured the absolute quantum yields of Quinine sulfate in a degassed 0.5 M H 2 SO 4 using the same instrument and the same sample cell in order to evaluate the accuracy of the absolute quantum yields observed by the instrument. The observed quantum yield values of 55-57% were in good agreement with the reported value of 55% 24 . Considering aforementioned results, we determined that the observed Φ values contain 5% error.
Emission lifetime measurements were recorded using a Hamamatsu Photonics, C4334 system equipped with a streak camera as a photo detector and a nitrogen laser for the 337 nm excitation.
The data of the emission data are summarized in Supplementary Table 1.
X-ray crystal structure determination. The single crystal X-ray diffraction measurements were performed using a Rigaku FR-E Superbright rotating-anode X-ray source with Mo-target (λ = 0.71075 Å), equipped with a Rigaku RAXIS VII imaging plate as a detector, at 200 K. The intensity data were collected via the ω-scan technique and empirically corrected for absorption. The structures of the complexes were solved by direct methods using SHELXS2014 25 . The structure refinements were carried out using full matrix least-squares (SHELXL2014) 25 3 ] 2+ cation, two thirds of the Cl − anion, and two and five sixths of the H 2 O molecules were crystallographically independent. All dppm ligands in the crystal were disordered at two positions, with an occupancy factor of 0.5. All of the benzene rings of dppm ligands were treated using AFIX 66 constraints and SIMU restraints. Some EADP restraints were applied to model the disordered chlorine and water molecules. All non-hydrogen atoms, except water molecules, were refined anisotropically. Hydrogen atoms were included in the calculated positions, except those of H 2 O molecules. For [1](OTf) 2 ·H 2 O, two [Au 2 (dppm) 3 ] 2+ cations, four OTf − anions, and two water molecules were crystallographically independent. All of the benzene rings of dppm ligands were treated using AFIX 66 constraints. All non-hydrogen atoms, except one OTf − anion, one benzene ring, and one water molecule, which were disordered into two parts with occupancy factors of 0.5, were refined anisotropically. Hydrogen atoms were included in the calculated positions. Three of the four OTf − anions were modeled to adopt ideal conformations using FRAG commands. These anions were treated using EADP or DELU restraints.
The PXRD pattern of [2]Cl 2 was indexed using the program DICVOL 26 to produce a monoclinic unit cell (a = 20.588 Å, b = 18.129 Å, c = 12.988 Å, β = 98.892˚, V = 4789.27 Å 3 ) with good figure of merit. The space group was assigned based on systematic absences as P2 1 /n. Unit cell and profile refinement were carried out using the Pawley method and led to a good fit (χ 2 = 3.87) for these lattice parameters and the space group. The structure solution was obtained using the simulated annealing method with the program DASH 27 . Two rigid groups: [Au 2 (dppm) 2 ], in which P-C phenyl were allowed to rotate, and two Cl − molecules, in asymmetric units and Z = 4 for space group P2 1 /n, were introduced using a constrained Z-matrix description. During annealing, 26 runs of 1 × 10 7 Monte Carlo moves were performed. The best structure obtained (Profile χ 2 = 24.45) was taken as the starting structural model for Rietveld refinement. The Rietveld refinement 28 of [2]Cl 2 was performed using the programs RIETAN-FP 29 and VESTA 30 , introducing disorder for each Cl − anion. Restraints but no constraints for all bond lengths were employed to maintain the molecular geometry. Atomic displacement parameters were refined isotopically. Absorption collection was applied using the RIETAN-FP Program. Final Rietveld refinement result: a = 20.5723(6) Å, b = 18.1157(5) Å, c = 12.975(3) Å, β = 98.900(2)°, V = 4777.5(2) Å 3 , R wp = 8.860% (R e = 10.850%), R p = 6.403%, R B = 6.775%, R F = 6.605%; 5,801 profile points (2θ range, 2 to 60°); 282 refined variables. The result is shown in Supplementary Fig. 9.
The PXRD pattern of [Au 2 Cl 2 (dppm) 2 ] was indexed using the program DICVOL 26 to produce a triclinic unit cell (a = 20.481 Å, b = 14.008 Å, c = 11.237 Å, α = 66.65°, β = 109.83°, γ = 125.98°, V = 2370.44 Å 3 ) with good a figure of merit. The space group was assigned as P-1 based on systematic absences. Unit cell and profile refinement were carried out using the Pawley method and led to a good fit (χ 2 = 3.10) for these lattice parameters and the space group. The structure solution was carried out using the simulated annealing method with the program DASH. Two rigid groups: [Au(dppm)]Cl, in which P-C phenyl were allowed to rotate in an asymmetric unit and Z = 2 for space group P-1, were introduced using a constrained Z-matrix description. During annealing, 26 runs of 1 × 10 7 Monte Carlo moves were performed. The best structure obtained (Profile χ 2 = 20.92) was taken as the starting structural model for Rietveld refinement. Absorption collection was applied using the RIETAN-FP Program. The Rietveld refinement of [Au 2 Cl 2 (dppm) 2 ] was performed using the programs RIETAN-FP and VESTA. Restraints but no constraints for all bond lengths were employed to maintain the molecular geometry. Atomic displacement parameters were refined isotopically. (R e = 10.504%), R p = 3.908%, R B = 1.505%, R F = 0.794%; 5,701 profile points (2θ range, 3 to 60°); 242 refined variables. The result is shown in Supplementary Fig. 10. DFT calculations. DFT calculations for the [1]Cl 2 ·H 2 O system were performed using the Gaussian 09 program 31 with the PBE0 functional. The def2-tzvppd basis set was applied for Au and P atoms in [1] 2+ . 6-311G(d) basis sets were applied for bridging C atoms between two P atoms in [1] 2+ , and 6-31G and STO-3G basis sets were applied for C and H atoms in phenyl rings in [1] 2+ , respectively. Additionally, 6-311+ G* basis sets for Cl − and H 2 O molecules were used. The single-point and time-dependent DFT calculations were carried out for [1] Cl 2 ·H 2 O. The initial structural parameters were taken from the single-crystal X-ray structure of [1]Cl 2 ·8.5H 2 O. The contour plots of HOMO-12, HOMO-11 and LUMO are shown in Supplementary Fig. 3. The calculated absorption spectrum of [1]Cl 2 ·H 2 O is illustrated in Supplementary Fig. 4. The major components in the calculated absorption spectrum were summarized in Supplementary Table 3. The optimized molecular structures in the singlet ground state and triplet excitation state are demonstrated in Supplementary Fig. 15.