Modulating the Optical Properties of BODIPY Dyes by Noncovalent Dimerization within a Flexible Coordination Cage

Aggregation of organic molecules can drastically affect their physicochemical properties. For instance, the optical properties of BODIPY dyes are inherently related to the degree of aggregation and the mutual orientation of BODIPY units within these aggregates. Whereas the noncovalent aggregation of various BODIPY dyes has been studied in diverse media, the ill-defined nature of these aggregates has made it difficult to elucidate the structure–property relationships. Here, we studied the encapsulation of three structurally simple BODIPY derivatives within the hydrophobic cavity of a water-soluble, flexible PdII6L4 coordination cage. The cavity size allowed for the selective encapsulation of two dye molecules, irrespective of the substitution pattern on the BODIPY core. Working with a model, a pentamethyl-substituted derivative, we found that the mutual orientation of two BODIPY units in the cage’s cavity was remarkably similar to that in the crystalline state of the free dye, allowing us to isolate and characterize the smallest possible noncovalent H-type BODIPY aggregate, namely, an H-dimer. Interestingly, a CF3-substituted BODIPY, known for forming J-type aggregates, was also encapsulated as an H-dimer. Taking advantage of the dynamic nature of encapsulation, we developed a system in which reversible switching between H- and J-aggregates can be induced for multiple cycles simply by addition and subsequent destruction of the cage. We expect that the ability to rapidly and reversibly manipulate the optical properties of supramolecular inclusion complexes in aqueous media will open up avenues for developing detection systems that operate within biological environments.


Materials and methods
All commercial chemicals were used as received unless stated otherwise. Solvents were dried according to standard procedures. Synthesis was carried out in oven-dried glassware. NMR spectra were recorded on a Bruker Avance III 400 MHz spectrometer, a Bruker Avance III HD 500 MHz spectrometer, or a Bruker Avance III 800 MHz spectrometer. Chemical shifts (δ) are given in ppm relative to residual proton solvent resonances (4.79 ppm for D2O and 7.26 ppm for CDCl3). 1 H DOSY measurements were performed on a Bruker Avance III 500 MHz spectrometer with temperature and gradient calibration prior to the measurements (the diffusion coefficient of the solvent was used as a calibration standard). A constant temperature of 298 K was maintained during all NMR measurements unless stated otherwise. Solution and solid-state UV-vis absorption spectra were recorded with a Shimadzu UV-2700 or a UV-3600 spectrophotometer. Emission and excitation spectra were recorded with a Shimadzu spectrofluorophotometer RF-5301 PC. Fluorescence quantum yields were determined on a Quantaurus-QY Absolute PL quantum yield spectrometer. Experimental details about the time-resolved fluorescence and absorbance spectroscopy can be found in Section 10. For details on the X-ray data collection and refinement, see Section 13. S3

Synthesis of cage 1
Coordination cage 1 was synthesized based on a reported literature procedure.     S7 BODIPY 3 was obtained as a byproduct in the synthesis of 8-acetoxymethyl-1,2,3,5,6,7-hexamethyl pyrromethene fluoroborate. Specifically, acetoxyacetyl chloride (0.87 mL; 8.1 mmol; 1.2 eq) was slowly added to a solution of 2,3,4-trimethyl-1H-pyrrole (1.478 g; 13.5 mmol; 2.0 eq) in dry CH2Cl2 (6 mL). During the addition, the solution warmed up and turned deep-red. Then, the mixture was refluxed for 1 h, cooled down, and poured onto n-hexane (100 mL). The solvent was evaporated to afford a deep-red solid. The solid was redissolved in dry CH2Cl2 (70 mL) containing 6.8 mL (39 mmol) of diisopropylethylamine. After having been stirred for 10 min at room temperature, BF3·OEt2 (7.2 mL; 58 mmol) was added dropwise and the solution was stirred for 1 h at the same temperature, turning violet. Then, the reaction was discontinued and the mixture was washed with a saturated aqueous solution of NaHCO3 (3 × 100 mL) and dried over Na2SO4. The solvent was evaporated and the resulting solid was purified by column chromatography (CH2Cl2/petroleum ether 1/1). BODIPY 3 was isolated as a dark-purple solid. Yield: 167 mg (0.55 mmol, 8%).
The peaks in the NMR spectra were assigned by analyzing 2D NMR spectra (not shown).    . ORTEP representation of the X-ray structure of BODIPY 3 (thermal ellipsoids at a 50% probability level). C, gray; N, blue; O, red; B, yellow; F, green (note that the X-ray structures of BODIPYs 2 and 4 have been reported previously 3,4

Encapsulation of BODIPYs 2, 3, and 4 within cage 1
General procedure for encapsulation: Cage 1 (9.0 mg, 2.8 µmol) was dissolved in D2O (0.7 mL). The resulting colorless solution was added to an excess (>5 eq) of solid 2, 3, or 4 (none of which is soluble in water) and the resulting suspensions were stirred overnight in the dark at ambient temperature. Then, the undissolved solids were removed by several cycles of centrifugation, resulting in clear, intensely colored solutions. Stirring for a longer time, heating, and applying sonication did not increase the encapsulation yields. The resulting solutions were stable in the dark at ambient temperature for at least several months (verified by NMR and UV-vis absorption spectroscopy).
The intense colorization of the aqueous solutions of 1 in the visible area allowed us to follow the uptake of each BODIPY dye by the cage over time. To this end, aliquots from the suspensions were taken at various times, subjected to repeated centrifugations, and analyzed by UV-vis absorption spectroscopy. Representative UV-vis spectra are shown in Figure S13. Based on the results replotted in Figure S14, we concluded that the uptake of all the BODIPY dyes is complete within ~10 h, with minor differences between dyes 2, 3, and 4.       Inclusion complex 22⊂1 was obtained in ~50% yield as determined by 1 H NMR spectroscopy. To help assign peaks due to 1 vs 2, as-prepared 22⊂1 was treated with extra free 1 (see Figure S16). Figure S16. 1 H NMR spectrum of as-prepared 22⊂1 (in the presence of free cage 1, which always remains partially unfilled) (bottom; red), after the addition of an extra 1 equiv of free 1 (center; green), and after the addition of an extra 2 equiv of free 1 (i.e., a total of 3 equiv of 1) (top; dark-blue) (400 MHz, D2O, 298 K). S14 Figure S17. 13

NMR characterization of inclusion complex 32⊂1
Inclusion complex 32⊂1 was obtained in a quantitative yield, as determined by 1 H NMR spectroscopy.       Inclusion complex 42⊂1 was obtained in ~62% yield, as determined by UV-vis absorption spectroscopy (see Section 9). The signals in the NMR spectra of 42⊂1 were significantly broader than those of the other two inclusion complexes, which made the analysis of 42⊂1 using 2D NMR methods challenging.    8. X-ray crystallography of complexes 22⊂1, 32⊂1, and 42⊂1 Table S1 lists the structural parameters of cage 1 (free and encapsulating BODIPYs 2-4). Crystals of inclusion complexes 22⊂1 and 42⊂1 contained a single type of species. Single crystals of 32⊂1 contained two alternating inclusion complexes with slightly different conformations.

Species
Pdax-Pdax distance Pdeq-Pdeq distance Angle at Pdax  Table S1. Structural parameters for empty 1 and 1 encapsulating BODIPYs 2-4. Pdax and Pdeq denote axial and equatorial palladium nodes, respectively; the Pdax-Pdax distance is defined as the distance between two axial palladium nodes. The Pdeq-Pdeq distance is defined as the average distance between two opposite equatorial palladium nodes of the cage. The angle at Pdax is defined as angle between two TIm ligand planes at the axial position. All calculations of distances and angles were performed using Mercury 4.2.0 software.  Table S2. Structural parameters for aggregated BODIPYs 2, 3, and 4 within single crystals and inside cage 1.
For the calculation of plane-to-plane distances between two neighboring BODIPY units, individual planes were defined and calculated as surfaces based on the central C3BN2 ring of each BODIPY. For calculating the center-to-center distances, centroids within the same C3BN2 core units were calculated. The transition dipole moment of the S1←S0 absorption is polarized along the long axis of the BODIPY chromophore. Thus, the slip angle, θ, between the two dipole moments in a closely packed dimer was calculated as the inner angle of the triangle C1-centroid1-centroid2 (see the main text, Figure 1e, bottom). All calculations of distances, angles, centroids, and planes were performed using Mercury 4.2.0 software. Figure S45. ORTEP representation of the X-ray structure of inclusion complex 22⊂1 (thermal ellipsoids at a 50% probability level). Hydrogens, anions, and solvent molecules were eliminated for clarity. Pd, yellow; C gray; N, blue; O, red; B, brown; F, green.

S31
S32 Figure S46. ORTEP representation of the X-ray structure of the first conformer of inclusion complex 32⊂1 (thermal ellipsoids at a 50% probability level). Hydrogens, anions, and solvent molecules were eliminated for clarity. Pd, yellow; C gray; N, blue; O, red; B, brown; F, green.
S33 Figure S47. ORTEP representation of the X-ray structure of the second conformer of inclusion complex 32⊂1 (thermal ellipsoids at a 50% probability level). Hydrogens, anions, and solvent molecules were eliminated for clarity. Pd, yellow; C gray; N, blue; O, red; B, brown; F, green.
S34 Figure S48. ORTEP representation of the X-ray structure of inclusion complex 42⊂1 (thermal ellipsoids at a 50% probability level). Hydrogens, anions, and solvent molecules were eliminated for clarity. Pd, yellow; C gray; N, blue; B, brown; F, green. Figure S49. Comparison of the X-ray structures of 22⊂1, 32⊂1, and 42⊂1 (note that whereas the X-ray structures of 22⊂1 and 42⊂1 contained only one type of species, that of 32⊂1 featured two slightly different conformations (in a 1:1 ratio). Only one of them in shown here; for the other conformation, see the CIF file).

22⊂1 32⊂1 42⊂1
S35 9. Steady-state optical properties of 2-4 and their inclusion complexes UV-vis absorption measurements were carried out on 0.04 mM solutions of inclusion complexes (the concentration in terms of cage units) at ambient temperature. The solution of free cage 1 is practically transparent in the visible range, with an absorption onset at ~380 nm. BODIPY 2 dissolved in MeCN (note that 2-4 are all insoluble in water) exhibits optical behavior typical of classical BODIPY dyes, namely, a sharp absorption band in the visible range centered at 491 nm (corresponding to the S1←S0 transition) with a blueshifted vibronic shoulder (see Figure S50A). A minor, broader absorption seen centered at 354 nm corresponds to the S2←S0 transition. Fluorophore 2 exhibits an intense green fluorescence with the emission band mirroring the absorption band. The emission band is centered at 502 nm, giving rise to a small Stokes shift (11 nm) characteristic of BODIPY dyes. BODIPY 3 has an intensive violet color, with a main absorbance band at 533 nm. The fluorescence quantum yield of 3 is much lower than that of 2 due to quenching effects associated with the CHO group. The excitation spectrum recorded on a solution of 3 does not follow its absorption spectrum (excitation maximum = 517 nm), with the emission band centered at the same wavelength as the absorption band (533 nm). We hypothesize that the excitation spectrum may originate from a minor yet highly emissive species, such as a rotamer or protonated 3. BODIPY 4 is red-purple with the main absorbance band at 548 nm. As reported previously, 4 it has a low fluorescence quantum yield, with a weak emission band at 595 nm. Note that the absorption bands of 3 and 4 are significantly broader than for 2, which is likely due to the bulkiness of the substituents at the meso position (CHO and CF3).  To investigate this process in more detail, we followed the process by UV-vis absorption and fluorescence spectroscopies. We tested several organic solvents, in which BODIPYs 2-4 are readily soluble (MeCN, acetone, DMF, 1,4-dioxane, DMSO, and MeOH). In a typical experiment, a 0.04 mM solution of 22⊂1 in 2 mL of water (the concentration in terms of cage units) was treated with 100 μL aliquots of an organic solvent. After the addition of each aliquot, the solution was thoroughly mixed and the spectrum was recorded. The aliquots were added until no further changes in the spectra (other than those due to dilution) were observed. We verified that in all the experiments, all the species (1, 2, and 22⊂1) remained dissolved in solution.
The release of 2 from 22⊂1 upon the addition of MeCN was accompanied by the appearance of a sharp, intense absorption band, characteristic of monomeric BODIPYs in solution (see Figure S53A). In addition, fluorescence spectroscopy showed a rapid increase in the emission intensity ( Figure S53C).
Concentration-corrected spectra show that the absorption coefficient of 2 within 1, ε2′, is substantially smaller than that of free 2 in solution (ε2; specifically, we found that ε2′ ≈ 0.45ε2). In other words, the release of 2 from the cage leads not only to a bathochromic shift but also to a substantial increase in the absorption intensity. Provided that the final solution contains only monomeric 2 in solution (i.e., no encapsulated 2), we can calculate, based on the molar absorption coefficient of 2 (~80,000 L mol -1 cm -1 ), 4 the percentage of 1 encapsulating the noncovalent dimer of 2. The result of this calculation (~50%) is in a very good agreement with the encapsulation efficiency determined by 1 H NMR.
Next, we investigated the release of 2 from 22⊂1 using other organic solvents. As Figure S53B shows, the release commences only after a certain volume percentage of a polar organic solvent has been added. We found that MeCN, dioxane, acetone, and DMF were similarly efficient in releasing the dye from the cage. However, higher volume fractions of DMSO and MeOH had to be added to reach the same level of release. In all cases, the final absorption spectra were identical to those of free 2 in the corresponding organic solvent, indicating that 2 was released from the cage in a quantitative fashion.
The same experiments were repeated with BODIPY 4 and inclusion complex 42Ì1 (Figures S53D-F). The results were analogous, except that, in contrast to BODIPY 2, the fluorescence emission of 4 is higher inside the cage than in organic solution, resulting in decreased fluorescence emission upon the addition of an organic solvent (see Figure S53F). Given the molar extinction coefficient of free 4, ε4′ ≈ 45,000 L mol -1 cm -1 , 4 the percentage of filled 1 at the onset of the experiment was determined as 62%.

Time-resolved optical properties of 2-4 and their inclusion complexes
For 2 and 22⊂1, femtosecond transient absorption measurements were carried out on a system based on a mode-locked Ti:sapphire oscillator (Spectra Physics Mai Tai SP). The oscillator produced a train of <120 fs pulses (bandwidth ~12 nm FWHM) with a peak wavelength at 800 nm, typically of 900 mW, corresponding to ~10 nJ per pulse. The weak oscillator pulses were amplified by a chirped-pulse regenerative amplifier (CPA, Spectra Physics Spitfire Ace). The pulses were first stretched to several picoseconds, then regeneratively amplified in a Ti:sapphire cavity, pumped by a pulsed Nd:YLF laser (Spectra Physics Empower 45) operating at 1 kHz. After the pulse was amplified and recompressed, its energy was about 5 mJ in a train of 1,000 Hz pulses. An independent pump pulse was obtained by pumping an optical parametric amplifier (Spectra Physics OPA-800CF) that produces 120-fs pulses tunable from 300 nm to 3 μm. One Watt of light amplified on Spitfire was used; the output power of the OPA varied between a few μJ and tens of μJ (depending on the chosen wavelength) at 1,000 Hz. In the reported experiments, the pump was turned to 440 nm and the optical densities of the samples in 1 mm and 2 mm optical path length cuvettes were kept between 0.2 and 0.4 at the excitation wavelength. Spectral corrections and analyses were performed using SURFACE XPLORER Pro (Ultrafast Systems) and Origin 9.1 (OriginLab) software. The nanosecond fluorescence was measured using a LP920 flash photolysis system (Edinburgh Instruments, UK). The system was pumped by a Nd:YAG third harmonic (355 nm) driving an OPO (Spectra Physics, USA, Quanta-Ray Lab series, and GWU Lasertechnik, Germany, VersaScan-355 midband). The spectra were collected on an intensified CCD camera (i-Star, Andor, UK), where the delay was determined by gating the image intensifier.  For 3, 32⊂1, 4, and 42⊂1, the samples were excited by a frequency-tripled Nd:YAG Q-switched laser, pumping an optical parametric oscillator (Ekspla NT342/C/3/UVE) with a pulse duration of 5 ns and a repetition rate of 10 Hz. The fluorescence was collected in the direction orthogonal to the laser incident beam using a 20 × 0.4 NA objective and spectrally filtered using a long pass filter onto a monochromator (Acton SpectraPro2150i), coupled to a photomultiplier (PMT) tube (Hamamatsu R10699). Transient emission measurements were recorded by a 600 MHz digital oscilloscope (LeCroy Wavesurfer 62Xs). Emission spectra were plotted by integrating over transient emission curve at each wavelength. Excitation pulse energy was measured by a pyroelectric sensor (PE9-C, Ophir Optronics). The same setup was used to acquire time-resolved spectra of 2 and 22⊂1, and results similar to those in Figure S62 were obtained.    For nanosecond-to-microsecond transient absorption measurements on 2 and 22⊂1, the same excitation beam as described on p. S40 was used in combination with an EOS-Sub-Nanosecond Transient Absorption Spectrometer (Ultrafast Systems, USA). The measurement was based on combining the ultrashort pump pulse with a white-light continuum generated by a photonic crystal fiber, thereby allowing for longer delays that are electronically triggered rather than using an optical delay line.

Competitive binding of BODIPY 2 vs azobenzene guests
We have previously demonstrated 1 that azobenzene (Azo in Figure S66) and tetra-o-fluoroazobenzene (F-Azo in Figure S66) can bind within the cavity of cage 1. Moreover, we roughly estimated the association constant of Azo, Ka = [Azo2⊂1]/[1]·[Azo] 2 , as ~10 9 M −2 . Although determining the Ka of F-Azo was not possible, we hypothesized that the binding strength would be higher owing to the presence C-F⋅⋅⋅H hydrogen bond interactions. 1 In order to obtain hints about the binding strength of the model BODIPY 2, we followed competitive binding of 2 vs Azo and 2 vs F-Azo in the presence of an equimolar amount of both guests with respect to cage 1 and in the presence of a three-fold excess of both guests with respect to 1. The blue trace in Figure S66A shows the UV-vis spectrum of an equilibrated mixture of cage 1, BODIPY 2, and Azo in a 1:1:1 molar ratio. When the amount of both guests was increased three times, absorbance at ~480 nm (due to encapsulated 2) increased slightly, whereas absorbance at 320 nm (due to encapsulated Azo) decreased ( Figure  S66A, red trace). This result indicates that in the presence of competition between 2 and Azo, BODIPY 2 is bound preferentially, i.e., its Ka is higher than 10 9 M −2 . In contrast, when the amounts of 2 and F-Azo were increased from 1 equiv to 3 equiv with respect to the cage, we observed a drastic decrease in the amount of encapsulated 2 (blue and red spectra in Figure S66B, respectively), indicating that F-azo is bound significantly stronger than BODIPY 2.

Reversible switching between H-and J-aggregates of BODIPY 4
The addition of 4 in MeOH (20 µL) to an excess (980 µL) of water leads to the formation of J-type aggregates. 4 These aggregates have a sharp absorbance band at >600 nm (the exact position, intensity, and shape of the band depend on the degree of aggregation) and a sharp emission band at ~630 nm (the very small Stokes shift is characteristic of J-aggregates). When this solution of J-aggregates of 4 (77 µM) was treated with an excess of cage 1 in water, the J-aggregates' absorbance band rapidly decreased as a result of the encapsulation of 4 within 1 as an H-type dimer. Figure S67 shows the changes in the UV-vis spectra of 4 during titration with aliquots of 1. The band due to J-aggregates can no longer be seen after the addition of 0.5 equiv of 1, whereas the H-dimer band increases until ~0.8 equiv of 1 has been added, in agreement with the ~60% encapsulation yield of 4 within 1. The J-to-H-aggregate transition can also be followed by fluorescence spectroscopy, which showed a decrease in the ~630 nm sharp emission band accompanied by a rapid increase in a broader, more intense band at ~585 nm ( Figure S68). Remarkably, the addition of 1 equiv of the cage broke the J-aggregates of 4 and induced the formation of 42⊂1 within only ~30 s. We hypothesized that the H-to-J-aggregate transition accompanying the encapsulation of 4 could be reversed by a controlled disassembly of cage 1, and that such disassembly could be achieved using aqueous KCN, which forms a strong complex with Pd 2+ . To verify that KCN can decompose the cage, a solution of 10 mg (3.14 μmol) of 1 in 0.7 mL of D2O in an NMR tube was treated with aliquots of 100 mM aqueous (D2O) KCN. After the addition of each aliquot of KCN (which induced the precipitation of a white solid), the NMR tube was shaken and an NMR spectrum was recorded.
Cyanide CNis a strong ligand for Pd 2+ ; therefore, the addition of KCN gradually displaces the TIm (1,3,5triimidazoylbenzene, "triimidazole") and TMEDA (N,N,N′,N′-tetramethylethylenediamine) ligands from the Pd 2+ centers, resulting ultimately in the formation of the [Pd(CN)4] 2complex. Interestingly, the addition of 1.0 equiv of KCN (with respect to the Pd 2+ centers) is sufficient to remove all the TIm ligands from the solution (TIm's aromatic protons are no longer detectable; Figure S69). Further addition of KCN replaces the bidentate TMEDA ligands; the NMR spectrum after the addition of 4.0 equiv corresponds to that of free TMEDA in water (two singlets at 2.55 ppm and 2.26 ppm, corresponding to TMEDA's CH2 and CH3 groups; Figure S69). Having established that KCN can readily disassemble cage 1, we studied its effect on inclusion complex 42⊂1. When a solution of 42⊂1 in water was titrated with aqueous KCN, the formation of J-aggregates of 4 was observed by UV-vis spectroscopy (see Figure S70). Notably, a substoichiometric amount of KCN was sufficient to restore the full intensity of the J-aggregate band. This process could also be followed by fluorescence spectroscopy (see Figure S71).  Figure S71. Changes in the fluorescence spectrum (λexc = 530 nm) of 42⊂1 in H2O (39 μM) following the injection of KCN (0.8 equiv with respect to Pd 2+ ) followed over time.
Finally, we hypothesized that the pronounced color change caused by treating 42⊂1 with KCN could be used to detect the presence of cyanide in water. To demonstrate a proof-of-concept system for cyanide detection, we soaked small pieces of agarose gels with aqueous 42⊂1 and treated them with KCN, among other inorganic salts (see Figure S72). We found that the only other salt causing visual changes similar to those induced by cyanide was thiocyanate, although it took significantly longer for color change to appear. Figure S72. (A) Two pieces of agarose gel (10×10×1 mm; prepared using 1 g of agarose and 50 mL of H2O) soaked with aqueous 42⊂1 (2 mL; 0.5 μM). (B) Left: a 10×10×1 mm piece of agarose gel soaked with aqueous 42⊂1 (2 mL; 0.5 μM) following immersion in a 760 μM KCN solution (2 mL) for 10 min. Right: a control sample not exposed to KCN. (C) A larger (60×50×1 mm) piece of agarose gel soaked with a 0.5 μM 42⊂1 and exposed to different salts (150 nmol each): KCN (1), KCl (2), NaI (3), NaSCN (4), Na2SO4 (5), NaNO3 (6), and AcOK (7). The photograph was taken 10 min after applying droplets of the salt solutions. For KCN, the color change was immediate; for NaSCN, it took several minutes. We have also tested HCl, Na2SO3, and NaOH (10 μL of 100 mM solutions, corresponding to 1 μmol); HCl and Na2SO3 did not cause any visual changes; NaOH induced the decomposition of BODIPY and rendered the gel colorless.
13. X-ray data collection and structure refinement Single crystals of inclusion complexes 22⊂1, 32⊂1, and 42⊂1 were obtained by slow evaporation of water from their respective aqueous solutions. Single crystals of 3 were obtained by slow evaporation of CH2Cl2 from a CH2Cl2/hexane solution mixture of 3. All crystals were coated in Paratone oil (Hampton Research) and mounted on MiTeGen loops. They were flash frozen in a liquid nitrogen stream of the Oxford Cryostream system. Data collection was performed under a stream of nitrogen at 100 K. The diffraction data of 22⊂1 and 42⊂1 were collected on a Rigaku XtaLAB PRO diffractometer using Cu-Kα radiation (1.54184 Å) and processed with CrysAlisPRO. The diffraction data of 3 and 32⊂1 were collected on a Bruker APEX-II Kappa CCD diffractometer using Mo-Kα radiation (0.7107 Å) and processed with Bruker SAINT. The structures were solved by direct methods using SHELXT. All non-hydrogen atoms were further refined by SHELXL with anisotropic displacement coefficients. Hydrogen atoms were placed in calculated positions and assigned isotropic displacement coefficients, U(H) = 1.2U(C) or 1.5U (C-methyl), and their coordinates were refined in riding mode. The SQUEEZE protocol of Platon was run on 22⊂1, 42⊂1, and 32⊂1. The crystal data and structural refinement are summarized in Table S3.  Table S3. Crystallographic data. *Derived from the crystal structure.