Noble Metal Complexes of a Bis-Caffeine Containing NHC Ligand

N-Heterocyclic carbenes (NHCs) have seen more and more use over the years. The go-to systems that are usually considered are derivatives of benzimidazole or imidazole. Caffeine possesses an imidazole unit and was already utilized as a carbene-type ligand; however, its use within a tridentate bis-NHC system has—to the best of our knowledge—not been reported so far. The synthesis of the ligand is straightforward and metal complexes are readily available via silver-salt metathesis. A platinum(II) and a palladium(II) complex were isolated and a crystal structure of the former was examined. For the Pt(II) complex, luminescence is observed in solid state as well as in solution.


Introduction
The use of N-heterocyclic carbenes (NHC) has gained significant attention in recent years [1][2][3]. This ligand class has developed tremendously since its discovery by Arduengo et al. in 1991 [4]. This is due to their electronic properties, most notably the strong σ-donating character, which makes them useful ligands for the coordination of transitionmetal ions [5,6]. Since then, complexes with NHC-type ligands have been intensively studied in the context of photophysical applications (e.g., with iron [7,8] or ruthenium [9]) as well as catalysis (e.g., with nickel [10], iron [11] or palladium [12]). Thereby, the ligand design has a strong influence on the properties of the resulting complexes, with steric and electronic properties having the most significant contributions [3,13,14]. The number of carbene sites, that surround the metal center, has been identified as one further relevant factor (i.e., the more carbene moieties, the stronger expressed are the observed effects) [15]. It has also been shown that the NHC scaffold can be varied; in this respect, imidazoleand benzimidazole-type systems represent the most prominent examples to be found in the available literature [6,16,17]. Possessing an imidazole-type substructure, caffeine has already experienced some use as an NHC ligand but the overall number of publications referring to this is still minor . For example, the antimicrobial and anticancer properties of caffeine-derived NHC complexes have been studied thoroughly [17,20,25,[30][31][32][33]. It has to emphasized that caffeine represents a promising starting point for the design of NHC-type ligands, due to its availability, low price and, most notably, green footprint compared to others. In here, the implementation of caffeine into a tridentate ligand framework, with pyridine as the central unit, is presented (Scheme 1). This particular strategy has already been used inter alia to prepare tridentate ligands with imidazole, benzimidazole and triazole units [6,[37][38][39][40][41][42]. Concerning Pt(II) complexes, the NHC-pincer approach, using CCC-or CNC-type ligands, has been intensively studied [43][44][45][46][47][48]. The methylene bridges in the presented ligand allow for a more relaxed and ideal octahedral or square-planar coordination geometry-a factor that has also been shown to have a significant impact [49]. Notably, the ligand design offers plenty of options to further build upon, since the 4-position of the pyridine unit can easily be modified, e.g., by halogen atoms and, thus, enabling subsequent C-C coupling reactions [42]. Moreover, substituents others than methyl could be introduced at the N-atoms of the purine-base moieties (e.g., long alkyl chains in order to mediate the solubility).

Results
The precursor (or protoligand) synthesis was accomplished via a straightforward twostep procedure (Scheme 2), in which commercially available 2,6-bis(bromomethyl)-pyridine (1) is first added to deprotonated theophylline (2). This reaction can be performed in neat DMF; however, the addition of water accelerates the deprotonation of the purine base. The overall reaction volume must be sufficiently large as the product swells enormously and turns the reaction mixture into a gel. Mechanical stirring is even required after some time has passed. The high viscosity of the reaction mixture also calls for an excess of theophylline to allow the reaction to proceed efficiently; the educt is later removed by copious washing to give 3 as a fluffy, white solid. Methylation of the imidazole unit is the next step. Mild methylating agents, such as Meerwein's salt, should be used at an ambient temperature to avoid over-methylation. Compound 3 is barely soluble in CH 2 Cl 2 but, nevertheless, the best results were obtained when the methylation was performed in this solvent and long reaction times were applied (i.e., up to 72 h, but at least 24 h). The yield of the two reaction steps is, owing to the solubility of the compounds, overall acceptable. NHCs can be prepared following different procedures. Today, the most commonly used methods rely on the in situ generation of the carbene by a strong base, followed by the addition of a suitable transition-metal salt to form the complex [13,50]. Recently, a slightly varied approach has been developed which uses a weaker base or a metal salt, that already bears a ligand that can act as a base itself. Firstly, the base route was tested. Specifically, the deprotonation step, which involved the addition of KOtBu or n-BuLi to a suspension of 4 in THF at -80 • C yielded a yellow solution. However, the addition of transition-metal salts to this solution-either directly or dissolved in a solution-did not afford the desired complexes. Notably, the reaction with Pd(OAc) 2 , which intrinsically carries its "own" base (i.e., the AcO − counterion), also did not give the desired complex. Subsequently, the second common route was tested. This route involves the formation of a Ag(I) precursor complex which can then be used for salt metathesis.
The silver complex 5 was generated by heating the protoligand 4 in a mixture of CH 3 CN/CH 2 Cl 2 with Ag 2 O at 50 • C for at least two hours and confirming the absence of the imidazolium C-H bond via 1 H-NMR spectroscopy (Supplementary Materials Figure  S10). The other protons experienced a shift of their signals in the NMR spectrum: The pyridine protons were shifted upfield by 0.1 ppm (C-H triplet in 4-position) and 0.4 ppm (C-H doublet in 3-position), respectively. The signal of the methylene groups was upfield shifted by 0.15 ppm; whereas, the signals of the CH 3 -groups were all shifted downfield with the imidazolium N-CH 3 featuring the largest shift of 0.08 ppm (the other two CH 3 -groups were only shifted marginally by 0.03 and 0.01 ppm, respectively). The 13 C-NMR reveals a large shift of the signal assigned to the Ag(I)-bound carbon when compared to 4 (i.e., 189 pm vs. 140 ppm; Figure S11). This chemical shift is indicative of the carbene structure, and, according to Huyn, the value ranks it as strong δ-donor [51]. Basically, the Ag(I) complexes can be either monomeric species, as depicted in Scheme 3, or form dimeric structures. The particular structure of these complexes is mainly dependent on the stoichiometry of the reaction (i.e., the number of Ag 2 O equivalents) as well as the used solvent. It has been reported that the solubility of Ag(I) salts in MeOH is lower when compared to CH 3 CN or CH 2 Cl 2 ; thus, the formation of monomers is favored in these solvents [52]. The structure of the complex, in turn, represents an important factor for the subsequent salt-metathesis reaction with respect to the required stoichiometry. In the present case, ESI-MS gave strong evidence for the monomeric structure of 5 ( Figure S3) Tandem MS, specifically travelling-wave ion-mobility MS (TWIM MS) coupled to gradient tandem MS (gMS 2 ) is highly suited for the deconvolution and characterization of superimposed ions, which were generated via ESI. This particular method enables one to distinguish between organometallic ions of identical m/z values by, in simple words, net-charge and shape [53]. However, this technique is demanding and its application to the present system seems disproportionate.
Complex 5 was subsequently allowed to react with different transition-metal salts in CH 3 CN. The metals that were tested during our studies were iron(II), nickel(II), palladium(II), platinum(II) and ruthenium(III) (the results obtained for iron, nickel and ruthenium are summarized in the Appendix A). The Pd(II) and Pt(II) complexes were prepared from PdCl 2 and K 2 PtCl 4 , respectively. In both cases, a yellow solution was obtained after the reaction with silver precursor 5 (Scheme 3). Both complexes, 6 and 7, were isolated after column chromatography using CH 3 CN/H 2 O/KNO 3 (40:4:1 ratio) as the eluent, and SiO 2 as the stationary phase, followed by precipitation with NH 4 PF 6 . The 1 H-NMR spectra of 6 and 7 revealed the absence of the imidazolium proton as well as a splitting of the signal of the methylene groups, suggesting a diastereotopic environment of those protons (see Figures 1, S12 and S14).
Compared to Ag(I) complex 5 and protoligand 4, the signals of the pyridine's protons revealed a significant downfield shift-the shift of 0.3 ppm and 0.6 ppm for the triplet and doublet, respectively, was almost identical for both complexes ( Figures S8 and S10). For the other signals, the Pd(II) complex 6 showed a slightly larger shift compared to its Pt(II) counterpart 7. Again, the signal of the CH 2 groups split into two doublets at 6.57 and 5.70 ppm for 6 and 6.40 and 5.42 ppm for 7. This was already observed for similar complexes that also feature a methylene bridge between the carbene and the central binding unit. Signal splitting is attributed to the helical twisted structure which can form due to the ligand's flexibility. This feature is absent when the carbene unit is directly attached to the pyridine ring [47,[54][55][56][57][58][59][60]. The overall splitting of the signals, however, is larger for 7 compared to 6 (1.01 ppm vs. 0.87 ppm). The methyl groups for both complexes are similar again, with the Pd(II) complex' imidazolium N-CH 3 group revealing a 0.05 ppm downfield shift compared to 7, with the other CH 3 groups having similar shifts. The diastereotopic protons were further studied by variable temperature coalescence experiments to identify the coalescence temperature at which the conformational change becomes faster than the NMR timescale. Temperature-dependent experiments were conducted up to 100 • C. For complex 6, at this temperature the signals just start to broaden; whereas, no change is observed for 7 (Figures S19 and S20). Thus, the coalescence temperature of the complexes is relatively high, probably around 120 • C for 6 and 140 • C or higher for 7. This difference, when compared to other published systems [61], could be explained by the bulkiness of the caffeine-similarly high conformational rigidity has also been observed for other large systems [60]. The shift of the pyridine protons into the downfield region is due to the deshielding of the protons. Compared to 5 and 4, the electron density in the pyridine ring is reduced due to the N-M bond which is now present. This effect is stronger for the proton in 4-position compared to those in 3/5-position due to the mesomeric connectivity with the nitrogen.
The 13 C-NMR spectrum of 6 reveals the carbene shift to be at 174 ppm; ( Figure S13) whereas for 7, the proton shift is 171 ppm. This means that in either case the ligand is only moderately σ-donating when compared to other systems [51]. High-resolution ESI-MS measurements confirmed the identity of the complexes and excluded the presence of bridged species of higher nuclearity.
The coordination of the platinum atom is square-planar with the two NHC groups in a trans-arrangement (C1-Pt1-C2 173.46(8) • ). The remaining two coordination sites are occupied by the central pyridine moiety and a chlorido ligand, being oriented linearly to each other (N1-Pt1-Cl1 178.50(5) • ). The two NHC planes are twisted out of the metal's coordination plane by 39.06(6) • and 46.97(6) • , respectively, and are oriented almost perpendicular to each other (angle between C 3 N 2 planes 84.04(8) • ). The PF 6 − counterion as well as the CH 3 CN molecules, which were present in the crystal structure, do not display any unusually close intermolecular contacts. The Pt-C bond lengths are virtually identical at 2.024(2) and 2.014(2) Å, which is in the typical range observed for Pt(II)-NHC complexes deposited in the Cambridge Structural Database (CSD) [62]. The observed values are very similar to what was published by Limbach et al. in 2011, who studied platinum complexes with CH 2 -bridged, tridendate imidazole NHC ligands [54]. However, Santra et al. could not provide a crystal structure for their benzimidazole derivative; these authors performed B3LYP DFT calculations instead, and the estimated bond lengths are also similar, though a bit larger than for the imidazole ones [61]. The C carbene -N bond lengths are in a range of 1.339(3) to 1.374(2) Å. These values are larger compared to those reported for the imidazole and benzimidazole derivatives, which is most likely due to the π back-bonding from the metal atom to the C-N π* orbital, which is stronger in this caffeine-derived system [63,64]. The crystallographic data are compiled in Tables S1-S6.

Photophysical Properties
Platinum(II) complexes sometimes feature room-temperature luminescence in the solid state [65,66]. This behavior can be related to metallophilic Pt···Pt interactions and/or π-π interactions within the crystal lattice; it has been shown that modifications of the ligand scaffold can tune the emission [65][66][67][68]. Therefore, the photophysical properties of compounds 4, 6 and 7 were studied.
There are a few notable aspects to be mentioned here. Firstly, while the solid complex is emissive when excited at 365 nm, the solutions only emit when the energy-rich 254 nm wavelength is used. This behavior can be related to the absorption spectra ( Figure 3). The precursor shows strong absorption at 258 and 270 nm; whereas, an additional band at ca. 317 is present for the complexes (Table 1). Secondly, solvatochromism is absent-at least qualitatively and at room temperature.  The solid-state emission of the complex is unlikely to stem from Pt···Pt interactions within the lattice, as the closest interatomic distance in the aforementioned crystal structure is ca. 7.2 Å, and thus much larger than the 3-4 Å that are usually observed in luminescent Pt-complexes [68,69]. A similar argument can be made for π-π interactions, as the distances are quite large between the two pyridine units. However, there seems to be an inverted stacking, i.e., the 5-membered ring is located on top of the caffeine's 6-membered ring with distances of 3.6-3.7 Å. The distance between caffeine's carbonyl moieties (O3/O1) and the pyridine nitrogen atom is approximately 3 Å.
The emission spectra of 4, 6 and 7 are very similar (Figure 4). The spectral features and the data derived from the crystal-structure analysis let us conclude that the observed emission of the complexes stems from the caffeine itself rather than from the metal centers. The red-shift of the emission of 6 relative to that of 7 is in good agreement with related series of complexes [70]. The maximum of the emission band is red-shifted and the overall absorption is broadened.
It was observed that the emission's wavelength of 4 depended on the applied excitation wavelength (Figure 5a). This so called "red-edge effect" was first discovered for quinine, as the common fluorescence standard, and has, since then, been observed for many different compounds [71]. The underlying mechanism, however, is not so clear and has been explained by different hypotheses [71][72][73][74]. However, pinpointing to one explanation is not so easy for this compound. This is due to the strong H-bonding and dipole interactions the caffeine moiety might be involved in. These interactions are apparent from concentration dependent NMR shifts and are related to the strong dipole interactions with the solvent due to the dipole moment of the caffeine as well as the free rotation around the methyl linkage to the pyridine.  This same dependence has also been observed for the 6 and 7 (Figure 5b,c), although in a much less pronounced fashion. This provides us insight into the mechanism of this phenomenon. The aforementioned temperature-dependent NMR studies already indicated the high importance of the methyl bridge. A similar effect is observed here. There is a trend from protoligand 4 to 6 to 7 regarding the red-shift of the emission; this trend qualitatively matches the coalescence behavior. Whereas, in 4 all groups may rotate freely, this is less so for the complexes due to the metal-carbon bond. While 6 still changes its conformation relatively fast, complex 7 is more rigid and less fluctuating. This behavior is expressed by the position of the emission band at different excitation wavelengths: For protologand 4, the observed ∆nm is 79; whereas, lower values were found for 6 and 7 (∆nm of 15.5 and 2.5 for 7, respectively; Table 1).
The solid-state emission of the compounds was evaluated via a plate reader; however, only 4 was emissive enough to collect data ( Figure 6). Overall the weak solid-state fluorescence of the complexes, in particular of 7, is most likely due to the quenching effects. The substitution of chloride by alkyne derivatives or heavy-atom ligands (e.g., AsPh 3 or SbPh 3 ) is expected to improve the photophysical performance significantly [62,69,75].  While it lacks the methyl bridge and the coordinating nitrogen, complex 8 is highly emissive in solution as well as in the solid state. It features absorption bands at 265, 323, 416 and 441 nm, the first two of which are comparable to the complexes presented here. They have been attributed to mixed metal to ligand charge transfer and ligand-centred (MLCT-LC) transitions [76]. It emits at 449 and 474 nm, which is comparable to 441 nm of 7. Complex 9 is a CNC-type complex which very closely resembles 7, only missing the methyl bridge and having imidazole units instead of caffeine. Unlike 7, 9 is highly emissive and features green to orange fluorescence, depending on the water content. The authors attributed this behaviour and the emission to the Pt-Pt dimers due to the proximity (3.5 Å).

Discussion and Conclusions
The Pd(II), Pt(II) and Ag(I) complexes of a new tridentate ligand, which contained two caffeine-derived NHC moieties, were synthesized. The overall straightforward approach is highly promising since the ligand scaffold can readily be modified via the pyridine-or the purine-based sub-units. It remains to be tested whether the platinum or the silver complex exhibit anti-tumor or anti-microbial activity [25,32]. Moreover, the Pd(II) complex is currently evaluated regarding its catalytic activity in, e.g., cross-coupling reactions [18]. The emission and absorption properties of 5, 6, and 7 are reported and a red-edge excitation shift has been observed. The synthesis and photophysical characterization of derivatives of Pt(II) complex 7 (with alkyne or heavy-atom ligands) is also the subject of ongoing research.

Conflicts of Interest:
The authors declare that they have no conflict of interest.
Sample Availability: Samples of the compounds are available from the authors.

Appendix A
Synthesis of Fe, Ni and Ru complexes. Initially, the synthesis of the homoleptic Fe(II) complex of 4 was attempted. For this purpose, FeBr 2 in THF was added, which lead to a brown solution. However, upon warming to room temperature, a brown precipitate formed which could not be re-dissolved. ESI-MS of this sample could corroborate the identify the desired complex, but further analysis was not possible due to solubility issues. The reaction with NiCl 2 (PPh 3 ) 2 yielded a yellow solution that turned brown over time. The 1 H-NMR spectrum revealed sharp signals and a shift of the protons compared to 5 and the imidazolium proton was absent ( Figure S16). When 4 was heated with RuCl 3 ·xH 2 O in the presence of silver oxide, the yellow suspension turned blue upon gentle heating, However, 1 H-NMR spectroscopy only revealed the ligand's signals. When air was introduced, the solution slowly turned black; this process was slowed down when the sample was kept at 5 • C. When attempting the synthesis from the silver salt 5, a brown precipitate was observed upon addition of the ruthenium salt. The 1 H-NMR spectrum again showed signals assignable to the ligand, and the imidazolium proton was recovered. This behavior might explain why the Fe(II) complex decomposes so rapidly, since its properties are expected to be similar to those of the Ru(II) complex.