Photochemical Resolution of a Thermally Inert Cyclometalated Ru(phbpy)(N–N)(Sulfoxide)⁺ Complex

Abstract

In this work a photosubstitution strategy is presented that can be used for the isolation of chiral organometallic complexes. A series of five cyclometalated complexes Ru(phbpy)(N−N)(DMSO-κS)](PF₆) ([1]PF₆-[5]PF₆) were synthesized and characterized, where Hphbpy = 6′-phenyl-2,2′-bipyridyl, and N–N = bpy (2,2′-bipyridine), phen (1,10-phenanthroline), dpq (pyrazino[2,3-f][1,10]phenanthroline), dppz (dipyrido[3,2-a:2′,3′-c]phenazine, or dppn (benzo[i]dipyrido[3,2-a,2′,3′-c]phenazine), respectively. Due to the asymmetry of the cyclometalated phbpy^– ligand, the corresponding [Ru(phbpy)(N–N)(DMSO-κS)]⁺complexes are chiral. The exceptional thermal inertness of the Ru–S bond made chiral resolution of these complexes by thermal ligand exchange impossible. However, photosubstitution by visible light irradiation in acetonitrile was possible for three of the five complexes ([1]PF₆-[3]PF₆). Further thermal coordination of the chiral sulfoxide (R)-methyl p-tolylsulfoxide to the photoproduct [Ru(phbpy)(phen)(NCMe)]PF₆, followed by reverse phase HPLC, led to the separation and characterization of the two diastereoisomers of [Ru(phbpy)(phen)(MeSO(C₇H₇))]PF₆, thus providing a new photochemical approach toward the synthesis of chiral cyclometalated ruthenium(II) complexes. Full photochemical, electrochemical, and frontier orbital characterization of the cyclometalated complexes [1]PF₆-[5]PF₆ was performed to explain why [4]PF₆ and [5]PF₆ are photochemically inert while [1]PF₆-[3]PF₆ perform selective photosubstitution.

Introduction

Since the clinical approval of cisplatin a great number of inorganic complexes with anticancer properties have been described, among which several ruthenium complexes have reached clinical trials. Currently, most research is focused on either compounds based upon the piano-stool Ru(II)η⁶-arene scaffold pioneered by the groups of Dyson and Sadler^1,2 or ruthenium(II) polypyridyl complexes, of which several (photoactive) candidates have been developed by the groups of Dunbar,³ Gasser,⁴ Glazer,⁵ Renfrew,⁶ Keyes,^7,8 Kodanko,^9,10 or Turro.¹¹ More recently cyclometalated analogues of these complexes have emerged as a new subclass of light-activatable anticancer complexes.^3,12,13 In this type of compounds, one nitrogen atom in a polypyridyl ligand has been replaced by a carbon atom, resulting in an organometallic metallacycle.¹⁴⁻¹⁷ As a consequence, cyclometalated compounds often show enhanced properties for chemotherapy or photodynamic therapy (PDT) than their noncyclometalated analogons.¹⁴ In particular, the lower charge of cyclometalated complexes leads to an increased lipophilicity, which in turn increases uptake in cancer cells¹⁸ and often leads to higher cytotoxicity¹⁹ toward cancer cells. In addition, cycloruthenated polypyridyl complexes have increased absorption in the red region of the spectrum, which is excellent for photochemotherapy. Whereas polypyridyl ruthenium complexes typically absorb between 400 and 600 nm,²⁰ a bathochromic shift is usually observed for cyclometalated compounds due to the destabilization of t_2g orbitals by the π-donating cyclometalated carbanionic ligand, potentially allowing activation of these compounds in the photodynamic window, (600–1000 nm) where light penetrates further into biological tissue.²¹ Although cyclometalation often leads to a significant decrease of the photosubstitution properties of ruthenium complexes, the group of Turro has reported two cyclometalated complexes, cis-[Ru(phpy)(phen)(MeCN)₂]PF₆ and cis-[Ru(phpy)(bpy)(MeCN)₂]PF₆, (phpy = 2-phenylpyridine), that are capable of exchanging their acetonitrile ligand upon light irradiation and are phototoxic in cancer cells.²²

Inspired by this work and following our investigation of caged ruthenium complexes with the general formula [Ru(tpy)(N–N)(L)]²⁺ in which L is a sulfur-based ligand and tpy = 2,2′;6′,2″-terpyridine, we herein investigated the preparation and properties of cycometalated analogues of this family of complexes where the carbanion is introduced in the tridentate ligand. Five complexes [1]PF₆–[5]PF₆ with the general formula [Ru(phbpy)(N–N)(DMSO-κS)]PF₆ with Hphbpy = 6′-phenyl-2,2′-bipyridyl and N–N = bpy (2,2′-bipyridine, [1]PF₆), phen (1,10-phenanthroline, [2]PF₆), dpq (pyrazino[2,3-f][1,10]phenanthroline, [3]PF₆), dppz (dipyrido[3,2-a:2′,3′-c]phenazine), [4]PF₆), and dppn (benzo[i]dipyrido[3,2-a:2′,3′-c]phenazine, [5]PF₆), respectively, were considered. Interestingly, by replacing one of the lateral nitrogen atoms of terpyridine in [Ru(tpy)(N–N)(L)]²⁺ by a carbon ligand, these ruthenium complexes become chiral, and using chiral monodentate sulfoxides should allow for separating their diastereomers.²³⁻²⁵ However, these cyclometalated complexes turned out to be substitutionally inert under thermal conditions, preventing displacement of DMSO in the racemic precursor. In order to achieve the resolution of [1]PF₆, it was therefore necessary to design a photochemical route. By investigating the photophysical properties and photoreactivity of these complexes, three of these complexes were found suitable for this approach, of which one was resolved using a chiral monodentate sulfoxide ligand.

Results

Synthesis and Crystal Structures

The first attempted route toward the synthesis of compounds [1]PF₆–[5]PF₆ (Figure 1), inspired by the report of Ryabov and co-workers,²⁶ consisted of the coordination of the terpyridine analogon Hphbpy to the ruthenium benzene dimer [(η⁶-C₆H₆)RuCl(μ-Cl)]₂. However, this approach afforded the intermediate species [Ru(phbpy)(MeCN)₃]PF₆ in a maximum yield of only 32% and proved to be difficult to scale up. Therefore, an alternative route depicted in Scheme 1 was developed. Starting from cis-[RuCl₂(DMSO-κS)₃(DMSO-κO)], the reaction of the bidentate ligand N–N = bpy, phen, dpq, dppz, or dppn was realized first, followed by cyclometalation using Hphbpy in the presence of a catalytic amount of N-methylmorpholine, affording the five compounds [Ru(phbpy)(N–N)(DMSO-κS)]PF₆ ([1]PF₆–[5]PF₆) as a racemic mixture of enantiomers in good yield (65–74%).

Figure 1.

Chemical structures of the complexes presented in this study. [Ru(phbpy)(N–N)(DMSO-κS)]⁺, where N-N = bpy, phen, dpq, dppz, or dppn.

Scheme 1. Reagents and Conditions.

(a) N–N = bpy in EtOH/DMSO (15:1), reflux, 86%; (b) HPhbpy, cat. N-methylmorpholine in MeOH/H₂O (5:1), reflux, 65%. For N–N = phen = 77% and 68%, N–N = dpq = 95% and 74%, N–N = dppz = 87% and 73%, NN = dppn = 96% and 65%.

Single crystals suitable for X-ray structure determination were obtained by slow vapor diffusion of ethyl acetate in dichloromethane for [1]PF₆, hexane in dichloromethane for [2]PF₆ and [3]PF₆, and toluene in DCM for [4]PF₆. All compounds crystallized in space groups having an inversion center, thus containing a (1:1) mixture of enantiomers. A selection of bond lengths and angles is shown in Table 1. As expected, the ruthenium centers in these compounds have a distorted octahedral geometry similar to that of their terpyridyl analogues.²⁷ Compared to [Ru(tpy)(bpy)(DMSO-κS)](OTf)₂ replacing the nitrogen within this scaffold with an anionic carbon atom has only a modest effect on the corresponding bond length, with Ru1−C1 in [1]PF₆ (2.043(2) Å) being almost as long as Ru1−N1 in its terpyridine analogue (2.079 Å).²⁸ Furthermore, compared to its noncyclometalated analogon the trans-influence of the carbon atom in phbpy^– results in an elongation of the Ru1–N2 bond length in [Ru(phbpy)(bpy)(DMSO-κS)]²⁺ (2.173(2) Å), whereas in [Ru(tpy)(bpy)(DMSO-κS)]²⁺ the Ru1–N3 length is 2.073(3) Å.²⁹ In contrast, the ruthenium–sulfur bond length is shorter in [1]PF₆ (2.2558(7) Å) than in [Ru(tpy)(bpy)(DMSO-κS)]²⁺ (2.282(1) Å) as a result of the increased electron density on ruthenium, leading to stronger backbonding into the π* orbital of the S-bound DMSO ligand. Overall, this electronic effect barely affects the angles between C1–Ru1–N3 for [1]PF₆ (158.67(12) Å) and N1–Ru1–N3 for [Ru(tpy)(bpy)(DMSO-κS)]²⁺ (157.92(8) Å), confirming their high structural similarity (Figure 2).

Table 1. Selected Bond Distances (Å) and Bond Angles (deg) for Complexes [1]PF₆, [2]PF₆, [3]PF₆, and [4]PF_6.

	[1]PF6	[2]PF6	[3]PF6	[4]PF6
Ru1–S1	2.2558(7)	2.2359(4)	2.2405(9)	2.210(3)
Ru1–C1	2.043(2)	2.041(3)	2.029(5)	2.030(1)
Ru1–N1	2.002(2)	2.004(2)	2.005(5)	2.019(7)
Ru1–N2	2.173(2)	2.164(2)	2.176(3)	2.180(1)
Ru1–N3	2.088(2)	2.110(2)	2.089(3)	2.094(3)
Ru1–N4	2.079(2)	2.091(2)	2.083(4)	2.071(4)
S1–O1	1.486(2)	1.489(2)	1.485(3)	1.501(6)
C1–Ru1–N2	157.92(8)	158.45(9)	158.5(2)	155.6(7)
N3–Ru1–N4	78.07(7)	78.67(7)	78.9(1)	78.2(1)
S1–Ru1–N4	96.25(5)	97.29(5)	96.6(1)	96.0(1)

Figure 2.

Displacement ellipsoid plots (50% probability level) of the cationic part of the crystal structure of [1]PF₆ (a), [2]PF₆ (b), [3]PF₆ (c), and [4]PF₆ (d). Hydrogen atom and counterions have been omitted for clarity.

Thermal Stability

With compounds [1]PF₆–[5]PF₆ in hand, we first attempted to obtain diastereomers by the thermal reaction of several chiral ligands as shown in Scheme 2 and summarized in Table 2 (entries 1–6). Heating [1]PF₆ and (R)-methyl p-tolylsulfoxide at increased temperatures (up to 120 °C) in DMF resulted in the formation of ruthenium(III) species, as observed by a green color, whereas lower temperatures only led to the recovery of starting materials. Further attempts to substitute the monodentate ligand with nonchiral ligands (entries 7 and 8) such as LiCl, pyridine, or acetonitrile also proved to be unsuccessful. This thermal inertness was highly unexpected, since terpyridine analogues of these complexes are known to readily exchange their monodentate ligand in similar or much milder conditions.³⁰ The only thermal substitution possible, observed with [4](PF₆)₂, was obtained by prolonged heating (16 h) in acetic acid, which resulted in the partial formation of [Ru(phbpy)(dppz)(AcOH)]⁺ as proven by mass spectrometry (found m/z 675.1, calcd. m/z 675.1). However, this species could not be isolated. Overall, the exceptional thermal inertness of the DMSO ligand in [1]PF₆–[5]PF₆ required the development of an alternative strategy for the resolution of this family of chiral complexes.

Scheme 2. General Approach for the Thermal Conversion of Complexes [1]PF₆, [2]PF₆, and [4]PF₆ with Different Monodentate Ligands L.

Table 2. Attempts of Ligand Exchange for [1]PF₆, [2]PF₆, and [4]PF₆.

entry	complex	ligand (L)	solvent	T (°C)	substitution	reaction time (h)
1	[1]PF6	(R)-methyl p-tolylsulfoxide (5 equiv)	DMF	120		16
2	[1]PF6	(R)-methyl p-tolylsulfoxide (5 equiv)	DMF	80		16
3	[1]PF6	(R)-methyl p-tolylsulfoxide (5 equiv)	EtOH 3:1 H2O	80		16
4	[4]PF6	biotin (20 equiv)	EtOH 3:1 H2O	80		16
5	[4]PF6	N-acetyl-l-methionine (20 equiv)	EtOH 3:1 H2O	80		16
6	[4]PF6	N-acetyl-l-cysteine methyl ester (20 equiv)	EtOH 3:1 H2O	80		16
7	[4]PF6	l-histidine methyl ester 2HCl (20 equiv)	EtOH 3:1 H2O	80		16
8	[2]PF6	LiCl (20 equiv)	EtOH 3:1 H2O	80		16
9	[4]PF6		MeCN	80		16
10	[4]PF6		pyridine	80		16
11	[4]PF6		acetic acid	80	yes	16

Photosubstitution

Replacing the DMSO ligands in these complexes was therefore attempted photochemically, monitoring the reaction using ¹H NMR. When a sample of [2]PF₆ was irradiated in acetonitrile with white light (hν ≥ 410 nm, Scheme 3), a clean photoconversion to a new species was observed, which was confirmed to be the acetonitrile adduct by mass spectrometry. As shown in Figure 3, the ¹H NMR spectra clearly demonstrate the formation of the single species [Ru(phbpy)(phen)(MeCN)]⁺ ([7]⁺) characterized by a doublet appearing at 9.88 ppm, while the doublet of the starting material at 10.49 ppm quantitatively disappeared. This photochemical behavior is comparable to the photosubstitution occurring in [Ru(tpy)(N–N)(X)]²⁺.³¹ In a similar fashion, the DMSO ligand in [1]PF₆ and [3]PF₆ could also be exchanged upon photoirradiation by deuterated acetonitrile to afford [Ru(phbpy)(bpy)(CD₃CN)]⁺ ([6]⁺) and [Ru(phbpy)(dpq)(CD₃CN)]⁺ ([8]⁺), respectively. However, [4]PF₆ and [5]PF₆ were not photosubstitutionally active, in contrast to the noncyclometalated analogons [Ru(tpy)(dppz)(SRR′)] and [Ru(tpy)(dppn)(SRR′)] (SRR′ = 2-(2-(2-(methylthio)ethoxy)ethoxy)ethyl-β-d-glucopyranoside)²⁷ that both exchange their thioether ligand upon light irradiation.²⁷

Scheme 3. Reagents and Conditions for the Synthesis of [11-A/C]HCO₂.

(a) hv ≥ 410 nm in CD₃CN. (b) i. (R)-Methyl p-tolylsulfoxide in MeOH, reflux, 16 h; ii. Reverse-phase HPLC (0.1% HCO₂H in MeCN/H₂O). (5% over two steps for [11-A]HCO₂, 4% over two steps for [11-C]HCO₂).

Figure 3.

Evolution of the ¹H NMR spectra of [2]PF₆ in CD₃CN (3.0 mg in 0.6 mL) upon irradiation with white light (>410 nm) from a 1000 W xenon Arc lamp fitted with 400 nm cutoff filter 1 cm from the light source at T = 298 K. Spectra were taken every 1 h, with t_irr = 7 h.

Resolving Diastereomers

The photoactivity of [1]PF₆–[3]PF₆ therefore allowed us to investigate separation of their enantiomers. [2]PF₆ was used as representative example. In a first attempt, [2]PF₆ was converted to [7]PF₆ using white light irradiation in deuterated acetonitrile (∼7 h). However, neither chiral HPLC nor crystallization using sodium (+)-tartrate allowed for resolving this intermediate. Instead, an alternative approach was used: racemic [7]PF₆ was allowed to react with an excess of enantiomerically pure (R)-methyl p-tolylsulfoxide in MeOH, affording a (1:1) mixture of diastereomers of (anticlockwise/clockwise) A/C-[Ru(phbpy)(phen)(R)-Methyl p-tolylsulfoxide)]PF₆, [11-A/C]HCO₂ (Scheme 3). Subsequent purification over a reverse phase HPLC column afforded [11-A]HCO₂ and [11-C]HCO₂ as their respective diastereomers in 9% yield (5% over two steps for [11-A]HCO₂ and 4% over two steps for [11-C]HCO₂ (Figure S6). ¹H NMR confirmed that fraction 1 corresponded to the R-C diastereomer, which is most apparent because of its more shielded α-proton of phen appearing at 10.64 ppm (Figure 4). Fraction 2 contained the R-A diastereomer, with a doublet appearing at 10.74 ppm (Figure 4). This deshielding effect on the α-proton on phen is most likely attributed to the interaction of the tolyl group with the bidentate ligand. This assumption was supported by NOESY experiments (Figure S8), which showed the absence of interaction between the methyl of the sulfoxide and phen, whereas a weak interaction was observed for [11-A]HCO₂ (Figure S8). Both [11-A]HCO₂ and [11-C]PF₆ are diastereomers and not enantiomers, so that specific rotation would not give any valuable information on their chirality. Circular dichroism (CD) was used instead to demonstrate they are related to the two enantiomers [2-A]⁺ and [2-C]⁺. The CD spectra of [11-A]PF₆ and [11-C]PF₆ in MeCN (Figure 5) displayed symmetrical curves typical for enantiomers, except in the region below 250 nm where the contribution of the chiral (R)-tolylsulfoxide ligand to the absorption becomes non-negligible.³² Around 450 nm, either positive or negative Cotton effects were observed for [11-A]PF₆ or [11-C]PF₆, respectively, which must originate from the ¹MLCT transitions. Theoretically, resolution of these complexes by performing blue light irradiation in acetonitrile may be tempting. However, photosubstitution is usually accompanied by racemization of the coordination sphere, so that thermal ligand substitution would be preferred.³³ This was however not possible due to the exceptional thermal stability of the sulfoxide cyclometalated complexes (see above) that prevented thermal displacement of the chiral sulfoxide to obtain isolated enantiomers of [A-7]⁺, [C-7]⁺, [A-2]⁺, or [C-2]⁺. However, the mirrored CD spectra of the diastereoisomers [11-A]HCO₂ and [11-C]HCO₂ provided a clear proof of the opposite chirality of these complexes.

Figure 4.

¹HNMR spectrum (850 MHz) of [11-C]PF₆ (top) and [11-A]PF₆ (bottom).

Figure 5.

Superposition of CD spectra of first fraction (black, [11-C]HCO₂) and second fraction (red, [11-A]HCO₂) eluted diastereoisomers. T = 293 K, c = 5 × 10^–5 M in MeCN.

Photophysical and Photochemical Characterization

The difference in photoreactivity between [1]⁺–[3]⁺ and [4]⁺–[5]⁺ was not straightforward to understand, and therefore a full photophysical characterization of the five complexes was carried out. The electronic absorption spectra (Figure S1) of these complexes show that they have a considerable bathochromic shift (∼40 nm, Table 3) and a significant broadening of their ¹MLCT band compared to [9]²⁺ (411 nm, Table 3). [4]⁺ and [5]⁺ have additional absorption bands around 370 and 410 nm, respectively. These are most likely π–π* transitions arising from the dppz and dppn ligand. The spectra of [6]⁺–[8]⁺ in acetonitrile also showed a shift of the ¹MLCT band of ∼50 nm compared to [10]²⁺. This bathochromic shift is common for cyclometalated ruthenium complexes^12,34 and is mostly ascribed to an increase in the energy of the highest occupied molecular orbital (HOMO, t_2g).¹²

Table 3. Lowest-Energy Absorption Maxima (λ_max), Molar Absorption Coefficients at λ_max (ε in M^–1 cm^–1), Photosubstitution Quantum Yields in Acetonitrile (Φ₄₅₀) at 298 K, ¹O₂ Quantum Yields (Φ_Δ) at 293 K, and Phosphorescence Quantum Yield (Φ_P) for [1]PF₆–[10](PF₆)₂.

complex	formula	λmax (εmax in M–1 cm–1)a	λem (nm)	ΦΔb	ΦPb	Φ450
[1]PF6	[Ru(phbpy)(bpy)(DMSO-κS)]PF6	476 (50 × 102)	786	3.2 × 10–2	1.6 × 10–4	4.1 × 10–5
[2]PF6	[Ru(phbpy)(phen)(DMSO-κS)]PF6	450 (57 × 102)	800	3.9 × 10–2	2.1 × 10–4	1.3 × 10–5
[3]PF6	[Ru(phbpy)(dpq)(DMSO-κS)]PF6	451 (83 × 102)	787	1.1 × 10–1	2.1 × 10–4	2.2 × 10–5
[4]PF6	[Ru(phbpy)(dppz)(DMSO-κS)]PF6	450 (84 × 102)	618	7.0 × 10–3	2.6 × 10–4	<10–6
[5]PF6	[Ru(phbpy)(dppn)(DMSO-κS)]PF6	450 (75 × 102)	672	<10–3	8.4 × 10–5	<10–6
[6]PF6	[Ru(phbpy)(bpy)(CD3CN)]PF6	525 (71 × 102)	n.d.	n.d.	n.d.
[7]PF6	[Ru(phbpy)(phen)(CD3CN)]PF6	503 (63 × 102)	n.d.	n.d.	n.d.
[8]PF6	[Ru(phbpy)(dpq)(CD3CN)]PF6	495 (119 × 102)	n.d.	n.d.	n.d.
[9](PF6)2	[Ru(tpy)(bpy)(DMSO-κS)](PF6)2	411 (75 × 102)	n.d.	n.d.	n.d.	1.6 × 10–2
[10](PF6)2	[Ru(tpy)(bpy)(MeCN)](PF6)2	455 (91 × 102)	n.d.	n.d.	n.d.

^aIn MeCN.

^bin CD₃OD.

Visible light excitation of ruthenium polypyridyl complexes typically leads to (1) ligand exchange, (2) phosphorescence and/or, (3) singlet oxygen generation. First, the ability of [1]PF₆–[5]PF₆ to exchange the DMSO ligand for a solvent molecule was quantified by UV–vis spectroscopy (Figure 6). As observed under white light irradiation (>450 nm), monochromatic blue light irradiation in acetonitrile (450 nm) left [4]PF₆ and [5]PF₆ unaffected, while [1]PF₆–[3]PF₆ converted to the acetonitrile complexes [6]PF₆–[8]PF₆ with clear isosbestic points (441 and 490 nm for [1]PF₆, 470 nm for [2]PF₆, and 455 nm for [3]PF₆) confirming the selectivity of the photoconversion. ESI-MS spectra taken after each reaction confirmed the formation of the acetonitrile photoproducts. The photosubstitution quantum yields (Φ₄₅₀) were found to be 4.1 × 10^–5 for [1]PF₆, 1.3 × 10^–5 for [2]PF₆, and 2.2 × 10^–5 for [3]PF₆, which is a thousand times lower than that measured for [Ru(tpy)(bpy)(DMSO-κS)]²⁺(Φ₄₅₀ = 1.6 × 10^–2). This decreased reactivity is most likely caused by the destabilization of the ³MC state due to increased electron density at the metal center brought by the strong σ-donor C atom, whereas stabilization of the ³MLCT leads to a larger energy gap between the ³MLCT and ³MC state, therefore making thermal population of the latter rather unlikely.³⁴ This interpretation is supported by previous work of the Turro group, who has demonstrated that the efficiency of the photosubstitution in sterically congested cyclometalated complexes is very low or absent.^12,22

Figure 6.

Time evolution of the electronic absorption spectra of [1]PF₆–[3]PF₆ and [9](PF₆)₂ in deoxygenated MeCN upon irradiation at 450 nm at T = 298 K. Spectra measured every 30 min (every 0.5 min for [9]PF₆). (a) [1](PF₆) t_irr = 16 h, [Ru]_tot = 5.78 × 10^–5 M, photon flux = 1.68 × 10^–7 mol s^–1. (b) [2](PF₆), t_irr = 23 h, [Ru]_tot = 6.08 × 10^–5 M, photon flux = 1.67 × 10^–7 mol s^–1. (c) [3]PF₆, t_irr = 16 h, [Ru]_tot = 4.06 × 10^–5 M, photon flux = 1.68 × 10^–7 mol s^–1. (d) [9](PF₆)₂, t_irr = 1 h, [Ru]_tot = 6.52 × 10^–5 M, photon flux = 5.54 × 10^–8 mol s^–1.

Second, emission maxima (λ_em) and emission quantum yields (Φ_P) for [1]PF₆–[5]PF₆ were measured in acetonitrile (Table 3). All compounds were found very weakly emissive with a slightly higher phosphorescence quantum yield compared to the polypyridyl complex [Ru(phbpy)(tpy)]⁺ (Φ_P = 5 × 10^–6).³⁵ The emission wavelengths found for [1]⁺–[3]⁺ are comparable to those of [Ru(phbpy)(tpy)]⁺ (786–800 nm versus 797 nm)³⁵ and are similar to complexes reported by the group of Turro and Sauvage.^12,35 For complexes [4]⁺ and [5]⁺ a blue-shifted emission (618 and 672 nm) was observed compared to [Ru(phbpy)(tpy)]⁺, which suggested a different type of excited state compared to [1]⁺–[3]⁺. Third, singlet oxygen quantum yields (Φ_Δ) were determined in deuterated methanol by measuring the emission of ¹O₂ at 1270 nm. Φ_Δ values lower than 0.04 were found for all complexes with the exception of [3]PF₆, which produced ¹O₂ with a photoefficiency (Φ_Δ) of 0.11. Interestingly, [Ru(phbpy)(dppn)(DMSO-κS)]⁺ did not show any singlet oxygen production, whereas its noncyclometalated analogues [Ru(tpy)(dppn)(CD₃OD)]²⁺ and [Ru(tpy)(dppn)(py)]²⁺ both have been demonstrated to be excellent ¹O₂ generators.^36,37 Overall, changing terpyridine into phenylbipyridine had great consequences on the photochemical and photophysical properties of this series of complexes. Therefore, to further understand the photophysical differences between complexes [1]⁺–[3]⁺and [4]⁺–[5]⁺ electrochemical studies and DFT calculations were carried out.

Electrochemistry and DFT

The electrochemical properties of complexes [1]PF₆–[7]PF₆ and [9](PF₆)₂–[10](PF₆)₂ were determined with cyclic voltammetry (Figure 7 and Table 4) to provide insight into the frontier orbitals of the cyclometalated complexes.³⁸ As summarized in Table 4, the cyclometalated DMSO complexes [1]PF₆–[5]PF₆ show quasi-reversible oxidation processes (I_pa/I_pc ≈ 1) with Ru^III/Ru^II couples near ∼+0.30 V vs Fc^0/+ whereas [9](PF₆)₂ showed an irreversible Ru^II → Ru^III oxidation at +1.23 V vs Fc^0/+. Although the irreversibility of the oxidation of [9](PF₆)₂ does not strictly speaking allow to analyze this oxidation potential to a HOMO energy level, for [1]PF₆–[5]PF₆ the low-lying, reversible oxidation suggests that the Ru(dπ)-based HOMO of the cyclometalated complexes is very high in energy, due to the π-donating character of the phbpy^– ligand.¹² As the irreversibility of the oxidation of [9](PF₆)₂ is attributed to linkage isomerization of DMSO from S-bound to O-bound,³⁹ cyclometalation also appears to prevents redox-induced linkage isomerization of the DMSO ligand, most likely due to the increased electron density on ruthenium. The quasi-reversible Ru^III/^II couple of the DMSO complexes [1]PF₆–[2]PF₆ also appeared at a higher potential (+0.30 V vs Fc^0/+) compared to that of the acetonitrile compounds [6]PF₆–[7]PF₆ (0.00 V vs Fc^0/+), which can be explained by the electronic effects of the monodentate ligand; κS-DMSO is a stronger π-acceptor than CD₃CN and therefore has a stronger electron withdrawing effect on ruthenium(II).⁴⁰ The ligand-based reductions for [1]PF₆–[3]PF₆ was found to have very similar energies, with quasi-reversible reductions around −2.0 V vs Fc^+/0, suggesting that these are phbpy-based. For [4]PF₆ and [5]PF₆ however the L^0/- appeared to occur at much less negative potentials (−1.4 V vs Fc^+/0 for [4]PF₆ and −1.2 V vs Fc^+/0 for [5]PF₆) due to the strong electron-accepting properties of the dipyridophenazine moieties. These first reductions being essentially reversible, the LUMO of these two complexes is dppz- or dppn-based, respectively.⁴¹ The experimental HOMO – LUMO gaps ΔE_exp, which can be approximated, for quasi-reversible redox couples, to the difference between E_ox and E_red (Figure 7, left), followed similar trends to the theoretical HOMO – LUMO gaps ΔE_th calculated by DFT (Table 4). ΔE_th were found very comparable indeed for complexes [1]PF₆–[3]PF₆ (ΔE_exp ≈ 2.2 V and ΔE_th ≈ 3.6 V) and much higher than that of [4]PF₆ and [5]PF₆ (ΔE_exp = 1.8 and 1.6 V, respectively, and ΔE_th = 3.13 and 1.86 V). The particularly low value of ΔE found for [4]PF₆ and [5]PF₆ suggested that the dppz and dppn ligands may generate low-lying excited states, which would explain the absence of photosubstitution with these two complexes.

Figure 7.

(a) Cyclic voltammograms of cyclometalated complexes [1]PF₆–[7]PF₆ and noncyclometalated complexes [9](PF₆)₂ and [10](PF₆)₂. Scan rate 100 mV s^–1, with the exception of [4]PF₆, [6]PF₆, [7]PF₆, and [9]PF₆ which were measured at 200 mV s^–1. L = DMSO-κS or CD₃CN. (b) Experimental (E_ox and E_red from cyclic voltammetry, in V vs. Fc^+/0, left axis) and calculated (from DFT, in eV, right axis) values of the HOMO energy, LUMO energy, and ΔE energy gap.

Table 4. Electrochemical Properties As Measured with Cyclic Voltammetry and Theoretical HOMO – LUMO Gaps Calculated by DFT^a.

		Eox (V)	Ipa/Ipc	Ered (V)	Ipc/Ipa	ΔEexp (V)c	ΔEth (eV)d
[Ru(phbpy)(bpy)(DMSO-κS)]PF6	[1]PF6	+0.30	0.99	–1.90	1.47	2.20	3.65
[Ru(phbpy)(phen)(DMSO-κS)]PF6	[2]PF6	+0.32	1.02	–1.89	1.11	2.21	3.65
[Ru(phbpy)(dpq)(DMSO-κS)]PF6	[3]PF6	+0.29	1.01	–1.87, −1.95	0.66, 2.23	2.16	3.57
[Ru(phbpy)(dppz)(DMSO-κS)]PF6	[4]PF6	+0.35	1.04	–1.43, −2.00	1.03	1.78	3.13
[Ru(phbpy)(dppn)(DMSO-κS)]PF6	[5]PF6	+0.36	1.05	–1.21, −1.82, −2.01	1.07, 1.52	1.57	2.86
[Ru(phbpy)(bpy)(CD3CN)]PF6	[6]PF6	0.00	1.00	–2.05	1.34	2.05
[Ru(phbpy)(phen)(CD3CN)]PF6	[7]PF6	+0.02	1.04	–2.05	1.38	2.07
[Ru(tpy)(bpy)(DMSO)](PF6)2	[9](PF6)2	+1.23b		–1.48	1.00	2.71
[Ru(tpy)(bpy)(MeCN)](PF6)2	[10](PF6)2	+0.92	0.95	–1.67	1.06	2.59	4.12

^aPotentials given vs. Fc^0/Fc⁺ in MeCN with 0.1 M [Bu₄N]PF₆ as supporting electrolyte. Complexes were measured at 298 K with a scan rate of 100 mV s^–1, with the exception of [4]PF₆, [6]PF₆ [7]PF₆, and [9]PF₆ which were measured at 200 mV s^–1.

^bE_pa.

^cΔE_th = E_LUMO – E_HOMO at the DFT/PBE0/TZP/COSMO level in water.

^dΔE_exp = E_ox – E_red,.

To confirm this hypothesis, density functional theory (DFT) calculations were performed for [1]⁺–[5]⁺ at the PBE0/TZP/COSMO level. The calculated HOMO energy, LUMO energy, and ΔE_th = E_LUMO – E_HOMO of the minimized geometries followed the same trend as the experimental values (Table 4 and Figure 7b). For [1]⁺ and [2]⁺ the LUMO was located on the phbpy^– ligand, for [4]⁺ and [5]⁺ it was localized on the dppz and dppn bidentate ligand, respectively (Figure S3–5), whereas for [3]⁺ empty orbitals localized both on phbpy^– and dpq were found close in energy and near the LUMO level. Thus, like for the terpyridine series,⁴¹ extending the conjugation of the bidentate ligand in the cyclometalated series [1]⁺ to [5]⁺ resulted in a strong stabilization of the LUMO in [4]⁺ and [5]⁺, and in a shift of its localization, from the tridentate ligand in [1]⁺ and [2]⁺ to the bidentate ligand in [4]⁺ and [5]⁺, with [3]⁺ as borderline species (Figure 7b). The strong stabilization of the LUMO in [4]⁺ and [5]⁺ generates low-lying excited states, most likely of ³π–π* character.

Discussion

Recent examples of the group of Turro have shown that complexes such as cis-[Ru(phpy)(phen)(CH₃CN)₂]PF₆ are as photoactive as their noncyclometalated counterparts, with a reported photosubstitution quantum yield (Φ_P) of 0.25 in dichloromethane.²² A more recent report by Albani et al. has shown that for [Ru(biq)₂(phpy)]PF₆ the phpy^– ligand increases the energy of the ³MC state, which in their case completely prevents photodissociation of the bidentate biq ligand.¹² In the family of complexes [1]⁺–[5]⁺ presented here (Figure 8), cyclometalation of the terpyridine ligand allows photoinduced ligand exchange for three of the five complexes ([1]⁺–[3]⁺), while it is absent in the more conjugated analogues [4]⁺ and [5]⁺. The photoreactivity of ruthenium complexes is result of a delicate interplay of excited states of different natures and energies. In [1]⁺, [2]⁺, and [3]⁺ the emission maximum was close to 800 nm, irrespective of the nature of the bidentate ligand, because the ³MLCT excited states must be located on the phbpy ligand. By contrast, in the more conjugated complexes [4]⁺ and [5]⁺ the emission maxima depend significantly on the bidentate ligand, with a higher energy (λ_em = 618 nm) for the less conjugated dppz complex, compared to dppn (λ_em = 672 nm, see Table 3). Two results are apparently contradictory: the higher energy of the emitting (³MLCT) excited states vs the very low calculated and experimental ΔE values in [4]⁺ and [5]⁺, compared to [1]⁺, [2]⁺, and [3]⁺. This contradiction suggests that the lower triplet states centered on dppz and dppn and arising from the photochemical population of the low-lying LUMO-like orbitals are not emissive; they are probably of ³π–π* character and centered on the phenazine moiety of the dppz or dppn ligand. The weakly emissive states, on the other hand, most likely of ³MLCT character, are higher in energy in [4]⁺ and [5]⁺ because they are centered on the bpy moiety of dppz or dppn, while in [1]⁺, [2]⁺, and [3]⁺ they are centered on the more conjugated phenyl-functionalized bipyridine ligand. All in all, the ligand photosubstitution reactions occur from metal-centered ³MC states, which are high in energy for [1]⁺–[5]⁺ due to the excellent σ-donor properties of the cyclometalated ligand and probably poorly dependent on the conjugation of the bidentate ligand. Due to the presence of their low-lying ³π–π* states, [4]⁺ and [5]⁺ cannot perform any photosubstitution, as nonradiative decay pathways are faster.⁴² For [1]⁺, [2]⁺, and [3]⁺ these ³π–π* states are much higher in energy, so that the photogenerated, low-lying phbpy-based ³MLCT states, in spite of the higher-lying ³MC states, still leads to photosubstitution, though at a significantly lower rate than in the terpyridine analogue [9]²⁺.

Figure 8.

LUMO orbitals for [Ru(tpy)(bpy)L]²⁺ ([9]²⁺) and for [1]⁺–[5]⁺ at the DFT/PBE0/TZP/COSMO level in water.

Chiral-at-metal complexes based upon ruthenium, iridium, or rhodium have been extensively investigated by the group of Meggers,⁴³⁻⁴⁵ Barton,⁴⁶ and others⁴⁷⁻⁴⁹ and have shown great promise in, e.g., asymmetric (photo)catalysis^50,51 or as anticancer drugs.⁵² To resolve these types of complexes, a classical method consists of coordinating an enantiomerically pure chiral auxiliary to the metal center, resulting in a mixture of diastereomers which can be separated in preparative scales using normal phase chromatography such as silica.⁵³ After separation, these diastereoisomers are typically treated with an achiral monodentate ligand of interest, thus resolving the two pure enantiomers. Other resolution methods involve direct recrystallization of enantiomers using chiral counterions such as Δ-TRISPHAT,^32,54−56 or separation of the enantiomers on chiral HPLC.⁴⁶ For [1]⁺–[5]⁺ these strategies could not be followed due on the one hand to the exceptional inertness of the coordination sphere and possibly to the very similar molecular shape of the cyclometalated vs. pyridyl side of the ruthenium-coordinated phbpy ligand. We hence relied on photochemical substitution to introduce a chiral sulfoxide ligand as resolving agent. The resulting diastereoisomeric ruthenium complexes [11-A]PF₆ and [11-C]PF₆ were inseparable on normal-phase silica. We therefore diverted to the use of reverse phase HPLC using 0.1% formic acid in the eluent. As a result, the isolation of the two diastereoisomers as their formate complexes was possible, but the presence of formic acid affected the overall yield (9%), most likely due to partial reprotonation of the cyclometalated ligand and subsequent (partial) degradation of the products. This is an issue that will be addressed in the future.

Conclusion

Replacing the terpyridine tridentate ligand in [Ru(tpy)(NN)L]²⁺ with phbpy has led to a new family of chiral-at-metal complexes [1]⁺–[5]⁺ with drastically altered thermal and photochemical properties compared to their polypyridine analogues. In particular, thermal substitution of the monodentate sulfoxide ligands becomes virtually impossible, while the ligand photosubstitution efficiency was reduced or even quenched due to the strong effect of cyclometalation on the energy of the HOMO and LUMO of the complexes. When N–N is a phenazine-based ligand ([4]⁺ or [5]⁺), the LUMO is based on the bidentate ligand and full quenching of the photoreactivity occurred, in great contrast to the photochemical behavior of terpyridine analogues such as [Ru(tpy)(dppz)(SRR’)]²⁺ or [Ru(tpy)(dppn)(SRR’)]²⁺ that undergo selective photosubstitution in water (Φ₄₅₀ = 0.02²⁷ and 0.00095,³⁶ respectively). The resolution of photosubstitutionally and thermally inert chiral cyclometalated complexes such as [4]⁺ and [5]⁺ will thus require strategies that still need to be developed. However, when N–N is bpy, phen, or dpq ([1]⁺–[3]⁺), selective photosubstitution of DMSO by acetonitrile remained possible. The ability of [1]⁺–[3]⁺ to exchange DMSO by acetonitrile upon visible light irradiation can be exploited, as demonstrated here with [2]⁺, to labilize the thermally inert achiral DMSO ligand and replace it in two steps by a chiral sulfoxide ligand, thus allowing the separation of the two chiral isomers [11-A]⁺ and [11-C]⁺. This works demonstrates that photosubstitution reactions can be useful for the resolution of chiral-at-metal organometallic complexes, which opens new synthetic routes toward catalytically or biologically active chiral organometallic complexes.

Supporting Information Available

The Supporting Information is available free of charge on the ACS Publications website at DOI: 10.1021/jacs.8b10264.

• Synthetic procedures for the synthesis of [1]PF₆-[8]PF₆, [11-A/C]HCO₂, [12]–[16]; quantum yield determination for [1]PF₆, [2]PF₆, [3]PF₆, and [9]PF₆; NMR irradiation experiments for [1]PF₆–[3]PF₆; electrochemistry experiments; CD spectra for [11-A/C]HCO₂; DFT structures for [1]PF₆–[5]PF₆; and the NOESY spectrum of [11-A]HCO₂ (PDF)

• CIF files for the studied compounds (ZIP)

Author Contributions

^§ These authors made an equal contribution to this work.

The authors declare no competing financial interest.