Exposing Mechanisms for Defect Clearance in Supramolecular Self-Assembly: Palladium–Pyridine Coordination Revisited

Spherical three-dimensional (3D) cages composed of palladium(II) and pyridyl ligands are a mainstay of supramolecular chemistry with demonstrated catalytic and optoelectronic applications. The widely reported self-assembly of these palladium-based cages exhibits sensitivity to the solvents, reagents, and/or reactants employed. This sensitivity, and the resulting inconsistency between synthetic protocols, hinders the development of desirable palladium-based cages. We have found that pyridyl ligand substitution—the rate-limiting step of self-assembly—is facilitated by endogenous supporting ligands derived from the solvents, reagents, and reactants employed in synthetic protocols of palladium- and platinum-based assemblies. Here, we present a systematic investigation combining 1H-NMR, electrospray ionization mass spectrometry (ESI—MS), and absorption spectroscopy to characterize the intermediates to support the mechanism of pyridyl ligand substitution on a model complex, M(py)2 (M = (N,N,N′,N′-tetramethylethylenediamine)palladium(II), py = pyridine), under simulated synthetic conditions for self-assembly. Our investigation exposes mechanisms for pyridyl ligand substitution, featuring intermediates stabilized by solvent, anion, or (in situ formed) alkoxide moieties. Interrogation of destabilizing agents (2,2,2-trifluoroethanol and tetra(n-butyl)ammonium chloride) reveal similar mechanisms that ultimately facilitate the self-assembly of coordination cages. These findings rationalize widely reported solvent and anion effects in the self-assembly of coordination cages (and similar constructs) while highlighting methodologies to understand the role of supporting ligands in coordination chemistry.


■ INTRODUCTION
Discrete supramolecular three-dimensional (3D) cages featuring palladium (II) coordination nodes have garnered significant interest in contemporary chemistry 1−18 due to their novel optoelectronic 19−28 and catalytic applications. 29−42 These Pd-based cages are synthesized by a self-assembly process that exploits the reversibility of coordination bonds between ditopic pyridyl ligands and Pd-containing coordination nodes to afford products bearing the minimum free energy (i.e., ΔG). 43−50 The intermediates of this synthetic process include defective frameworks, wherein two ligands occupy a single ligand site during self-assembly (Scheme 1, orange). 51−60 Importantly, the process of defect clearance via substitution of a coordinated ditopic ligand (i.e., pyridyl ligand substitution) is the rate-limiting step in the self-assembly of coordination cages. 51 The literature proposes an associative mechanism for pyridyl ligand substitution, where an incoming ligand coordinates to an existing (four-coordinate) Pd complex. The result is the formation of a five-coordinate transition state, which ejects a ligand to yield a four-coordinate complex. 51 This mechanism is supported by theoretical studies, 61,62 but does not account for the significant steric hindrance incurred that can inhibit defect clearance within the framework (Scheme 1, orange). 63−66 The nature of this steric encumbrance is exacerbated by the geometric constraints imposed on pyridyl ligands that are incorporated into the surrounding rigid framework (Scheme 1, blue). 67 Reported self-assembly protocols for supramolecular cages utilize a diversity of Pd 2+ salt precursors and reaction solvents, highlighting the significance of judicious pairing of reagent/ reactant/solvent combinations in achieving desirable products with good synthetic yields. 68 While many studies have detailed the effect of ditopic ligand design on the topology of product cages, 1−16,63−70 few have considered the effect of specific solvents, reagents, or reactants (i.e., Pd 2+ precursor). Hiraoka et al. reported that the use of coordinating solvents (specifically, MeCN) and elevated temperatures (343.15 K) improved the yield of cuboctahedral Pd 12 L 24 coordination cages by facilitating pyridyl ligand substitution and promoting defect clearance. 71 Fujita et al. demonstrated that 2,2,2trifluoroethanol (TFE) as a co-solvent (80% v/v in DMSO) improved the formation of similar Pt 12 L 24 cuboctahedral cages based on platinum (II) metal centers, acting through the destabilization of coordination bonds. 72 Recently, we reported that trace quantities of chloride (Cl − ) or the addition of carboxylic acids facilitate the self-assembly of both Pd 12 L 24 and Pt 12 L 24 . 69,73 We propose that these additive reagents catalyze the supramolecular self-assembly of Pd n L 2n cages through a common mechanism, namely, by facilitating the elementary step of pyridyl ligand substitution at Pd coordination nodes. Thus, a realization of reactant or reagent effects on the thermodynamics of pyridyl ligand substitution should generate a rationale that improves the synthesis of many Pd-based coordination cages.
In this work, we engage in a systematic investigation of pyridyl ligand substitution of a model complex M(py) 2 (M = (N,N,N′,N′-tetramethylethylenediamine)palladium(II), py = pyridine). By employing variable-temperature (VT) 1 H-NMR and 1D-exchange spectroscopy (EXSY), we determine the rates and activation energies (E a ) for pyridyl ligand substitution under different reagent and reactant conditions. Specifically, we demonstrate that solvents, counterions, and additives (Cl − and TFE) enable different mechanisms of pyridyl ligand substitution facilitated by endogenous supporting ligands. These mechanisms are further evidenced by electrospray ionization mass spectrometry (ESI�MS), computational, and VT absorption spectroscopy studies of the intermediate M(py)(L) complexes (where L is the endogenous supporting ligand, e.g., MeCN). We infer that analogous pyridyl ligand substitution mechanisms likely occur during the self-assembly of Pd-based coordination cages. This leads to improved synthetic yields observed in previous reports, 71−73 as rate-limiting defect clearance within the framework is highly favored. 51−60 Importantly, this mechanism relies upon smallmolecular entities as supporting ligands, circumventing the steric hindrance evoked by the presence of axially-coordinated pyridyl ligands. 61 ) or dimethylsulfoxide-d 6 (DMSO). These samples were analyzed by 1 H-NMR to observe the speciation of mono-pyridyl M(py)(L) (where L is a supporting ligand such as a solvent or anion) and bis-pyridyl M(py) 2 complexes ( Figure 1).
In addition to M(py) 2 and free py (Figure 1, resonances α′ and α, respectively), a second complex is observed (Figure 1, resonance α″) that was identified as M(py)(L) (where L is a supporting ligand) by diffusion-ordered spectroscopy in DMSO (DOSY, Figure S1). This species is observed only in samples containing a lower excess of py ([py] = 58.3 mM). The formation of M(py)(L) in the presence of an excess of py indicates that entropic penalties of forming M(py) 2 are not countered by the enthalpic gains from bond formation between Pd 2+ and py. 74 In contrast, a greater excess of py ([py] = 167.3 mM) results in the quantitative formation of M(py) 2 .
While the literature proposes that the association of free py limits pyridyl ligand substitution in Pd-based complexes, 51 we observe the formation of M(py)(L) is suggestive of a supporting ligand-dependent or dissociative mechanism ( Figure 1b). These results are supported by CSI�HRMS measurements (Figure S6), and 1 H-NMR spectra ( Figure S39) that indicate the presence of M(py)(NO 3 − ) (i.e., L = NO 3 − ) under these conditions. Therefore, we propose that NO 3 − acts Scheme 1. Defect Clearance from Pd 2+ Frameworks as an endogenous ligand to encourage the formation of M(py)(L) complexes that are amenable to further substitution.

EXSY Analysis of py Substitution in M(py) 2 Complexes in Coordinating and Noncoordinating Solvents.
The dissociation of M(py) 2 affords an M(py)(L) complex (where L is a supporting ligand) and free py (eq 1).
With one-dimensional (1D) gradient-edited EXSY, the excitation of the α-proton of the py ligand within M(py) 2 ( Table 1, α′) was used to quantify the evolution of peaks for the products of py substitution ( Figure 2). This approach is sensitive to the peaks arising from chemical exchange allowing for direct measurement of rate. 75 The excitation of α-protons in coordinated py (Figure 2, green, α′) leads to the evolution of peaks corresponding to free py ( Figure 2, blue, α) and M(py)(L) (Figure 2, pink, α″). Using a standard equation (eq 2), 75 we computed the rate constant (k, s −1 ) of chemical exchange or reaction from the integrated areas of the excited peak (α′ as I excited ) and an evolved peak (α or α″ as I evolved ) arising over the mixing time of the EXSY experiment (T m , s).
From the data in Figure 2, we computed the rate constants for the production of free py (k py ) and M(py)(L) (k M(py)(L) ) afforded by the dissociation of M(py) 2 complexes (Table 2). When our model complex is formed with a lower excess of py ([py] = 58.3 mM), k py and k M(py)(L) are identical, which is expected as these form from the same complex (Table 2, 86 A rapid association suggests that pyridyl ligand substitution is limited by the dissociation of the pyridyl ligand, which we observe directly as k py . This py-dissociation-limited mechanism is further supported by the invariance of k py to changes in py concentration in both coordinating (DMSO, k py = 0.035 s −1 ) and noncoordinating (MeNO 2 , k py = 0.011 s −1 ) solvents. However, the difference in these rates points to the role of endogenous ligands (i.e., L) and the importance of solvent coordination (L = DMSO) in the pyridyl ligand substitution process. Thermochemical Analysis of Pyridyl Ligand Substitution with M(py) 2 Complexes in the Presence of Different Anions. The use of VT−EXSY allows the measurement of pyridyl ligand substitution rates across a range of temperatures. Fitting these rate data with the Arrhenius equation (eq 3) affords both activation energy (E a ), as a characteristic thermodynamic barrier for pyridyl ligand substitution at M(py) 2 , and a preexponential factor (A) representing environmental contributions.
Crucially, E a reflects the transition state enthalpy of the ratelimiting step of the specific reaction mechanism, which is   Inorganic Chemistry pubs.acs.org/IC Article independent of concentration or entropic contributions. Therefore, these measured E a values reflect both the mechanism and facility of pyridyl ligand substitution under different reactant and reagent conditions. To distinguish differences in pyridyl ligand substitution from solvent or anion selection, we determined the E a of py dissociation in M(py) 2 complexes in both MeNO 2 and DMSO solvents, where the initial complexes were prepared with a range of  51 We propose that this mechanism of solvent-facilitated pyridyl ligand substitution occurs with all coordinating solvents (Figure 3b, L = S, generally), leading to their favorable use in self-assembly protocols. 68 When a noncoordinating solvent (i.e., MeNO 2 ) is used, the E a of pyridyl ligand substitution of M(py) 2 complexes becomes sensitive (i.e., dependent) on the anion used ( Figure 3, right columns). As E a varies only with the relative enthalpy of the five-coordinate transition state, these anion effects do not originate from the solubility of the anion. This leads us to surmise that each anion affords a unique intermediate complex, M(py)(A), via a five-coordinate transition state followed by ejection of a py ligand (Figure 3b, L = A).
The formation and subsequent degradation of M(py)(A) results in the substitution of a py ligand at the M(py) 2 center. Stabilization of the five-coordinate transition state, M(py) 2 (A), leads to a favorable E a (i.e., minimum E a ). We found that the anion electrostatic volume (V esp ) correlated to the measured E a values ( Figure S36), indicating that the steric volume (or charge density) of an ion serves to limit its efficacy as a supporting ligand.
Interestingly, NO 3 − is smaller than DMSO (V esp = 10.41 vs 14.71 Å 3 ) and stabilizes the transition state more effectively (E a = 11.61 vs 12.08 kcal mol −1 ). Despite these physical and thermodynamic metrics, solvent-facilitated ligand substitution is preferred. Presumably, this preference is due to the sheer abundance of DMSO compared to NO 3 − (14.1 M vs 34.6 mM, respectively, ca. 400 equiv), but this also does not account for any elevation in local concentrations that arise from ion pairing in solution. From the reports of Hiraoka and co-workers on transient supramolecular cage formation complexes (Scheme 1), 71 we considered that solvation may determine the mechanism of pyridyl ligand substitution.
MD Simulation of Solvation Effects on Anion-Pair Formation. The ion pairing that arises from the electrostatic attraction between M(py) 2 and NO 3 − may be mediated by intermediary polar-solvent molecules providing a basis for observed changes in ion interactions under different solvent conditions ( Figure 3). 85 Stronger ion pairing increases the local concentration of NO 3 − near the M(py) 2 complex and therefore elevates the collisional frequency of contact between them. Unfortunately, DOSY and 1 H-NMR are not suited to directly observe ion pairing interactions between M(py) 2 complexes and NO 3 − in solution. While measurement with 14 N-, or 15 N-NMR may be possible, we could not obtain sufficiently well-resolved spectra at their respective frequencies with our instrumentation (data not provided). Instead, we developed MD simulations to observe how solvent interaction attenuates ion pairing (Scheme S2). From these MD trajectories (1 μs), we visualized the volumes traversed by NO 3 − and quantified NO 3 − ion pairing using computed diffusion-and distance-based metrics (Figure 4).
Our MD simulations directly show that ion pairing between NO 3 − and M(py) 2 is sensitive to solvent conditions, and the increasingly favorable solvation by DMSO/water enables NO 3 − to traverse across larger volumes around the complex during the simulation (Figure 4b,c) compared to MeNO 2 (Figure 4a). The mobility of NO 3 − about the M(py) 2 complex was measured as the quotient of the diffusion constants for the two species (D NOd 3 /D M , unitless), 85 where a ratio of 1.00 observed for MeNO 2 shows minimal translational freedom  Figure S33a,b). Results are additionally provided numerically in Table S1. The electrostatic volumes (V esp , Å 3 ) of anions were determined by fitting DFT-computed electron densities ( Figure S35).

Inorganic Chemistry pubs.acs.org/IC
Article between the ions (Figure 4a). The NO 3 − anions proved to be increasingly mobile in DMSO and water leading to diffusivity ratios of 1.09 and 1.16, consistent with solvent separation (Figure 4b,c). 85 This effect is also visualized by the distribution of Pd−NO 3 − distances (Figure 4d), demonstrating that coordinating or protic solvents, in general, diminish the frequency of contact between NO 3 − and M. These results evidence that solvents drive mechanistic preferences by impairing anion-facilitated pyridyl ligand substitution. Since this effect may be independent of a solvent's capacity for coordination, this effect may be exploited to control or halt self-assembly to isolate kinetically trapped assemblies.
Thermochemical Investigation of Coordinating, Noncoordinating, and Protic Solvents on Pyridyl Ligand Substitution. As a solvent or co-solvent, TFE is uniquely employed in the self-assembly of Pd n L 2n cages to improve synthetic yields. 72 However, the effect of other protic solvents (e.g., water, MeOH, EtOH) on self-assembly or pyridyl ligand substitution is unclear. As protic solvents are weakly coordinating 81 and readily solvate NO 3 − (Figure 4c,d), it is uncertain if pyridyl ligand substitution is facilitated by endogenous ligands under these conditions or unaffected by solvent or anion choices via a dissociation-limited mechanism. Therefore, we determined the characteristic E a of pyridine dissociation using our VT−NMR approach to gain insight into the role of solvent proticity in ligand substitution ( Figure 5).
Our VT−NMR measurements reveal that aprotic solvents have a lower E a value (11.1−12.5 kcal mol −1 ) for pyridyl ligand substitution compared to the studied protic solvents (15.2− 21.3 kcal mol −1 ). The lower E a of aprotic solvents suggests improved defect clearance and is consistent with their frequent   S20b, S27b−S30b, S30c) in noncoordinating (orange), coordinating (purple), and protic solvents (green), shown with reference pK a values. 82 Activation energies were determined by the Arrhenius equation using EXSY-determined ligand substitution rates obtained over a range of temperatures appropriate for each solvent (i.e., from 300 K to within 5 K of solvent boiling point, see Figure S33c). These results are provided numerically in Table S2.
Contrasting behavior was observed for pyridyl ligand substitution in protic media, with large differences in the E a apparent that correlated to the respective proton association (i.e., pK a ) of the solvent. Previous reports propose that TFE acts as an acid, protonating the M-coordinated py and disrupting the py-Pd 2+ to yield M(py)(L) and pyH + . 72 However, quantum-chemical modeling of the H + association to the bound py with GFN2-xTB 87 suggests that electrostatic repulsion between the Pd 2+ center of M(py) 2 and H + occurs with a higher energy barrier than the direct dissociation of py ( Figure S37). Therefore, we propose that the dissociation of a proton from solvent affords an alkoxide (i.e., RO − ) that acts as the anionic supporting ligand to facilitate pyridyl ligand substitution (Figure 3b, L = RO − ). This proposed mode of action rationalizes the use of TFE (pK a = 11.4) to promote the formation of large coordination cages by providing a reservoir of 2,2,2-trifluoroethanolate (TFEO − ) as supporting ligands for pyridyl ligand substitution. 72 Characterization of Intermediate M(py)(L) Complexes Using VT Absorption Spectroscopy. To support our mechanistic findings, we developed an experimental protocol to directly distinguish solvent or anion-coordinated intermediates that are inaccessible by current approaches; this approach is detailed in Scheme S2. This protocol uses a series of absorption spectra for samples of M(py) 2 complexes in different solvents, starting at 300 K and followed by 5−10 elevated temperatures, to reveal regions of linearly increased absorption (Scheme S2). By comparing these temperaturedependent absorption differences to TDDFT calculation results, we could directly interpret the spectra and distinguish solvent or anion-coordinated complexes during pyridyl ligand substitution in situ ( Figure 6).
The experimental absorption spectra feature narrow peaks within the UV region, suggesting the occurrence of metalcentered d−d transitions (Figure 6, black traces). 84 The TDDFT-calculated vertical transitions and extrapolated spectra for the modeled M(py)(L) intermediate complexes ( Figure  6a−c, inset) are in good agreement with the observed spectra, enabling qualitative comparison between the two (Figure 6, blue traces). To test the validity of our approach, we applied the same TDDFT methods to oppositional models (i.e., M(py)(NO 3 − ) in DMSO), finding that these oppositional vertical transitions were a complete mismatch between coordinating solvents or anions (Figures S40−S44). This approach has a limited ability to distinguish M(py)(NO 3 − ) and M(py)(RO − ), which have nearly identical computed vertical transitions (Figures S45−S47). However, the consistency between observed spectra and TDDFT  71 However, the complexes we observe feature endogenous ligands rather than the 3chloropyridine that were employed in the quantitative studies of cage formation. 71 As the majority of synthetic protocols lack additive supporting ligands (i.e., 3-chloropyridine), supramolecular cage formation complexes similar to those observed In our EXSY measurements, the excitation of α′ led to the observation of α and α″ peaks ( Figures S31c−S32c), establishing the rate of py and M(py)(A) production during the measurement. In the absence of additives, these quantities would be the same ( Figure 2); however, in their presence, pyridyl ligand substitution may be facilitated by either the endogenous ligand (Figure 7a, pink) or the additive anion (Figure 7a, yellow). Under these conditions, the py production in excess of M(py)(A) corresponds to the endogenously facilitated mechanism, allowing us to quantify the contribution of these two mechanisms to pyridyl ligand substitution ( Figure  7b).
Surprisingly, we also observed the dissociation of py from M(py)(A) complexes in our VT−EXSY spectra ( Figures  S31b−S32b). We presume that the dissociation of py affords a neutral complex M(A) 2 , which could not be observed in 1 H-NMR due to spectral overlap with either TFE or tetra(nbutyl)ammonium. Similar to the interconversion between M(py) 2 and M(py)(A) (Figure 7a, pink), the interconversion between M(py)(A) and M(A) 2 results in pyridyl ligand substitution by a secondary mechanism (Figure 7a, purple). Interestingly, the rate of this secondary py substitution at M(py)(A) complexes was similar to or greater than the primary anion-facilitated mechanism at M(py) 2 (Figure 7b, yellow). Remarkably, when TFE was used as a solvent, the majority of observed ligand substitution occurred through this additional mechanism (Figure 7b, purple). In contrast, when Cl − is used with DMSO as solvent, both Cl − and DMSO independently facilitate pyridyl ligand substitution. While DMSO-facilitated pyridyl ligand substitution is hampered by additive Cl − (Figure 7b, pink), the total rate is enhanced�by up to 52% at 350 K�from the additive effect of these mechanisms.
We propose that the formation of analogous complexes by Cl − /TFEO − at Pd 2+ coordination nodes facilitate defect clearance in the formation of Pd-based coordination cages and similar frameworks (Scheme 1), rationalizing a common mechanism for the reported effects of destabilizing additives in self-assembly. 71−73 Importantly, the self-assembly of multicomponent constructs (e.g., large coordination cages) 1−60 would benefit from similar intermediates as they facilitate ratelimiting defect clearance. 51−60 ■ CONCLUSIONS In this work, we have developed methods to investigate pyridyl ligand substitution as the rate-limiting process in the selfassembly of coordination cages. Using VT−EXSY and VT absorption spectroscopy, we demonstrated that pyridyl ligand substitution is facilitated by endogenous supporting ligands originating from solvents, counterions, or additives present in the reaction. We found that the polarity, proticity, and coordinating ability of the solvent used dictate the mechanism  (V esp ). Similarly, with protic solvents, in situ formed hydroxide or alkoxides lead to an E a dependent on solvent pK a . These anion-facilitated mechanisms of pyridyl ligand substitution are also found when 2,2,2-trifluoroethanol (neat, as a solvent) or tetra(n-butyl)ammonium chloride (as additive) are employed, rationalizing the reported effects of these reagents toward enhancing synthetic yields of Pd-and Pt-based coordination cages. Importantly, our thermodynamic findings have implications to improve the self-assembly of spherical coordination cages and similar frameworks. The defect clearance by pyridyl ligand substitution is enhanced by: (a) the use of coordinating solvents (e.g., DMSO), (b) pairing a small, coordinating anion with a nonpolar, noncoordinating solvent (e.g., NO 3− and MeNO 2 ), or (c) the use of destabilizing additives such as Cl − or TFE. Our findings may be similarly leveraged to improve cage stability or induce the formation of kinetically trapped species by removing endogenous ligand sources to increase the E a of pyridyl ligand substitution.
Method, experimental data (including NMR, VT absorption, and CSI-HRMS spectra), and computational model details (PDF) ■ AUTHOR INFORMATION Corresponding Author