Mechanism of the Photodecomposition of Stable Triarylmethyl Radicals

ABSTRACT

Luminescent radicals, the vast majority of which are derivatives of tris(trichlorophenyl)methyl (TTM), are of significant recent interest because of the unique photophysical properties of the doublet excited state. Though they show high chemical stability, most trityl radicals show very poor photostability, which hinders their application as magnetic, optical and quantum‐related materials. In this work, we use density functional theory to study the mechanism of photodegradation of TTM. We isolate the photodecomposition products and characterize them via mass spectrometry, NMR, EPR, UV‐Vis absorption spectroscopy, cyclic voltammetry (CV), and X‐ray crystallography. We show that the reaction proceeds by a 5‐electron electrocyclization followed by an unusual 1,8‐sigmatropic chloride shift, affording two fluorenyl radicals, which slowly oxidize and hydrolyze to form semiquinone products. We carefully examine the reported photostability of >80 substituted triarylmethyl radicals and demonstrate that other common triarylmethyl radicals, including benchmark luminescent derivatives with the highest photostability, the carbazole‐appended TTMs, photodecompose through the same cyclization mechanism, and thus the DFT‐calculated activation energy of cyclization can be used to guide the design of photostability in new luminescent triarylmethyl radicals.

Photolysis of luminescent chlorinated trityl radicals was studied with density functional theory and the decomposition products were isolated and characterized. It proceeds by a 5‐electron electrocyclization then a 1,8‐sigmatropic chloride shift, forming non‐emissive, less stable fluorenyl radicals. Calculated activation energies predict the experimental photostability of trityl radicals, including the benchmark carbazole‐substituted TTM.

1. Introduction

Stable organic radicals—molecules with unpaired electrons—are important materials for magnetic, spintronic, energy storage and bioimaging applications [1, 2]. Most recently, luminescent radicals have attracted tremendous attention for optoelectronic and emerging quantum technologies. The doublet spin state of excitons in radical materials endows some unique behaviors not accessible in closed‐shell molecules [3, 4]. The absence of dark triplet states in radical emitters enables 100% quantum efficiency in electroluminescent devices [3, 5, 6], while the combination of magnetism with luminescence enables their application as optically‐addressable qudits [7] and bioimaging materials [8, 9]. Virtually all of these applications have been developed with polychlorinated triarylmethyls (TAMs, Figure 1a,b) [4, 10, 11, 12]. In particular, TTM derivatives substituted with a carbazole or arylamine group are among the best‐explored and show some of the greatest performance as optical materials, with high quantum yields (>50%) and emission wavelengths ranging from pure red to the near‐IR region [4, 13, 14, 15, 16].

FIGURE 1.

Structures of (a) TTM, PTM and (b) selected derivatives with measured improved photostability versus TTM (comprehensive review of the literature, with >80 examples, in Scheme S1 and Table S1). (c) Putative photodecomposition mechanism of PTM proposed by Fox et al.

While such chlorinated TAMs show remarkable chemical stability [17], their poor photostability is one of the greatest barriers to future commercialization: under ambient light, the lifetime of PTM (perchlorotriphenylmethyl, Figure 1a) and TTM in solution is mere minutes [18]. Addressing this problem has been a subject of a significant research effort: More than eighty TAM derivatives with improved photostability have been reported in the last two decades (Figure 1b, complete list in Table S1). The explored strategies involve the substitution of chlorine in the para position with diarylamines (notably, carbazole derivatives), cyano [19] or aryl [20, 21] groups, or substitution of one or more of the trichlorophenyls on TTM with a pyridyl [22, 23] or carbazole [24, 25] heterocycle. Some of the most photostable TAMs known, developed by the groups of Li and Friend, bear a 3‐(N‐arylcarbazole) group and were reported to have photodecomposition half‐lives ca. 80 000 times longer than TTM [5, 26]. The only other known derivatives with comparable photostability are pyridyl TAMs developed by Nishihara and Kusamoto, which show photodecomposition half‐lives of up to 10 000–50 000 times longer than TTM [27, 28]. However, the photostability of luminescent radicals is still inferior to that of many closed‐shell dyes, and its improvement remains an important goal.

In spite of the extensive effort toward the design of photostable TAMs, the photodecomposition mechanism of these radicals has received no attention in the recent literature, though during the preparation of this manuscript, we became aware of the mechanistic study of photodecomposition in another TTM‐carbazole derivative by Hosokai et al. [29]. In 1968, a photoinduced ring closure of PTM with formation of fluorenyl radical PPF was proposed by Luckhurst and Ockwell (Figure 1c) [30]. However, in 1974, Bogatyreva and Buchachenko studied PTM photodecomposition with EPR in various solvents and concluded that it proceeds via hydrogen atom abstraction from the solvent forming the corresponding triarylmethane PTM‐H [31]. In 1987, Fox et al. used a combination of transient absorption and photoacoustic spectroscopies to study the photolysis of PTM radical in carbon tetrachloride [32]. They isolated the fluorenyl radical PPF as the major product of photodecomposition, reported time‐resolved spectra of intermediates, and proposed their plausible structures. This paper remains the most comprehensive and solid analysis of TAM photodecomposition. Nonetheless the reported evolution of the UV‐vis absorption spectra appears inconsistent with PPF as the main product. Furthermore, the applicability of this mechanism for TTM and the newer, more photostable derivatives has never been examined.

In this work, we use DFT and experimental techniques to study the photodecomposition mechanism of TTM and its derivatives for the first time. DFT calculations show that the 5‐electron photocyclization occurs irreversibly with a small barrier of 6.3 kcal/mol, followed by an unprecedented 1,8‐sigmatropic chloride shift, which leads to two fluorenyl radical products by loss of Cl₂ or HCl. DFT calculations on other common triarylmethyl radicals predict that the photocyclization mechanism should still be operative in these species. Finally, we show that photodecomposition of TTM‐3PCz—the benchmark of stable and efficient radical emitters [5]—proceeds through the same photocyclization mechanism as TTM.

2. Results and Discussion

2.1. Computational Study of Photodecomposition of TTM

We began our study with an analysis of the excited state of TTM. TD‐DFT calculations (uM06‐2X/6‐31G(d), gas phase) predict that the optimized first excited state D₁ lies 59.9 kcal/mol (2.59 eV) above TTM D₀. The D₁ transition is mostly comprised of a HOMO→SUMO component and could be qualitatively described as an electronic transition from one of the aryl rings to the central carbon. This transition shortens the C*–C^Ph bond from 1.47 to 1.41 Å, lengthens the C ^ipso –C ^ortho bond from 1.41 to 1.48 Å, and increases the spin density on the aryl ring, especially on the ortho and para positions, from ∼5.1% to ∼8.4% (Figure S9).

To examine the feasibility of TTM's decomposition into H‐TTM, as suggested by Bogatyreva et al. [31], we calculated the transition state of hydrogen atom abstraction from chloroform on the D₁ surface and found it prohibitively high (ΔG ^‡ = 42.4 kcal/mol, Figure S15). Conversely, the pericyclic reaction of TTM may be more facile in the excited state than the ground state because in the D₁ state there is an increase in spin density on the ortho carbons, which would facilitate the formation of a C–C bond between adjacent rings.

While the Woodward‐Hoffmann rules reliably predict the outcome of pericyclic reactions for closed‐shell systems, they are not applicable for radical electrocyclizations [33, 34, 35]. We thus computed both of the two possible products of photoinduced conrotatory and disrotatory ring closure, Int1 (anti) and Int1a (syn). As expected, Int1a has significant steric strain and lies 12 kcal/mol above TTM in the excited (D₁) state (Figure 2 and Table S3). In contrast, Int1(D₁) is about 5 kcal/mol below TTM(D₁), suggesting that the photocyclization is energetically favorable and sterically controlled to proceed exclusively through conrotatory ring closure. The transition state TS1 on the D₁ surface between TTM and Int1 lies only 6.3 kcal/mol above TTM (Figure 2). For comparison, the corresponding TS1’ state along the D₀ potential energy surface is calculated at 49.8 kcal/mol, which is in line with the very high temperature (>300°C) required for the thermally‐promoted cyclization reaction of PTM [36].

FIGURE 2.

(a) Mechanism for photodecomposition of TTM via dechlorinative cyclization. (b) Energy diagram [37] for the photocyclization of TTM (uM06‐2X/6‐31G(d), gas phase). Green arrows connect molecules on the D₁ and D₀ surfaces. Dashed lines connect states where the transition state (TS) was not calculated.

There are several possible fates to Int1(D₁) in the reaction mechanism. It can (i) deactivate (radiatively or non‐radiatively) to the ground state Int1(D₀) or (ii) form Int2 directly in the ground state (S₀) via scission of the C_sp3–Cl bond. This bond is weakened by the steric strain of Int1 (bond length of 1.97 Å; cf. 1.82 Å in CH₃Cl) while the product of its scission, Int2, is stabilized by rearomatization of one phenyl ring. The formation of Int2(S₀) from Int1(D₁) is exergonic by 26.9 kcal/mol due to the potential energy surface crossing to the ground state. A relaxed coordinate scan of Int1 on the D₁ surface elongating the C_sp3–Cl bond shows no apparent barrier (Figure S11). However, instead of moving away from the molecule, the chlorine atom drifts toward a different carbon of an adjacent phenyl ring in a rare [38, 39] 1,8‐sigmatropic shift forming Int1b. The feasibility of this unusual sigmatropic shift is defined by the geometry of Int1, where the chlorine atom is already within the van der Waals distance (3.0 Å) of the carbon atom onto which it migrates (Figure S10). It is also plausible that, instead of a sigmatropic shift, the chlorine atom dissociates from Int1, forming Int2 in the ground state (S₀). A relaxed coordinate scan of C–Cl dissociation from Int1(D₀) suggests a very small barrier (∼6 kcal/mol, Figure S12) for such a process, even in the gas phase. We could not find stable excited state (D₁/S₁) geometries for either Int1b or Int2; the calculations were nonconvergent due to instability of the C_sp3–Cl bond. Most likely, these intermediates are formed in the ground state, by crossing the potential energy surface from D₁ to D₀/S₀. The formation of the final photodecomposition product Cl₇PF (similar to PPF from perchlorinated PTM [32] isolated by Fox et al.) from Int2 is expected to be rapid and irreversible, plausibly assisted by the Cl radical formed in the same solvent cage [40] in the previous step. Indeed, the calculated transition state of chlorine‐assisted Cl‐abstraction from Int2 is only 5.6 kcal/mol (Figure S16) and the reaction is exergonic by 24.5 kcal/mol.

We next examined the possible transformations of the deactivated (ground state) Int1. Calculations on the D₀ surface reveal that the activation barrier for the rearrangement of Int1 to Int1b (TS2) is only 6.4 kcal/mol and the process is exergonic by 25.0 kcal/mol. The predicted intermediate Int1b can, in principle, form a second product, Cl₈PF, via the loss of HCl. The formation of Cl₈PF is thermodynamically favored versus Cl₇PF by 28.1 kcal/mol due to the formation of the strong H–Cl bond. This process is expected to be bimolecular, presumably assisted by chlorine (the by‐product of Cl₇PF formation), but extremely facile. The calculated activation barrier of chlorine‐assisted H‐abstraction from Int1b is only 0.8 kcal/mol (Figure S17).

2.2. Experimental Verification of TTM Photodecomposition Products

To experimentally validate the above mechanism, we performed photolysis of TTM and isolated the decomposition products. A 0.01 M solution of TTM radical in dry, anoxic cyclohexane was irradiated with a 370 nm 45 W LED lamp (ca. 70 mW/cm², details in Supporting Information) for ∼20 min. Reaction completion was monitored by fluorescence loss and thin‐layer chromatography (TLC), which showed the formation of a new, non‐emissive product with a slightly lower retention factor (R_f). This product was isolated at 86% yield by flash chromatography, affording a pink solid that was characterized by high‐resolution mass spectrometry (HRMS), UV‐vis absorption and EPR spectroscopies (Figure 3) and assigned as fluorenyl radical Cl₇PF. Notably, when the reaction was repeated at 0.02 M, two products were isolated: Cl₇PF (38% yield) and a new product, with a slightly higher R_f (25% yield), which was identified as Cl₈PF (Figure 3). The formation of the latter was predicted above (Figure 2) to occur via Int1b, through the elimination of HCl. This elimination is, most likely, a bimolecular process, explaining why Cl₈PF is not observed at lower concentrations. We tested the headspace of the reaction tube for Cl₂ and HCl gases and found both were present upon completion of the reaction (Figure S3).

FIGURE 3.

(a) TLC plate (silica; hexanes eluent) of TTM and the crude mixtures after photolysis at different concentrations. (b) Absorption spectra of TTM and the isolated fluorenyl radical photoproducts in CH₂Cl₂ (TD‐DFT calculated spectra in Figure S13 and Table S4). Inset: expanded view of Cl₇PF's D₁ band. (c) Mass spectra of the fluorenyl radicals, APCI negative mode. (d) CV of TTM and isolated fluorenyl radicals (arrows indicate starting point and direction of scan). (e) X‐band EPR spectra of the fluorenyl radicals and TTM in CH₂Cl₂.

The UV‐vis spectra of the isolated fluorenyl radicals (Figure 3b) show absorption bands between 450 and 600 nm (Cl₇PF: Ɛ = 2800 M⁻¹ cm⁻¹ at 547 nm) and extremely weak, broad D₁ transitions around 600–1000 nm (Cl₇PF: Ɛ = 140 M⁻¹ cm⁻¹ at 715 nm, Figure S18), even weaker than the symmetry‐forbidden D₁ transition of TTM (Ɛ = 1030 M⁻¹ cm⁻¹ at 542 nm [41]), consistent with predictions by TD‐DFT (Table 1). The distinct vibronic structure of the fluorenyls’ absorption bands can be attributed to the rigid structure of the fluorene ring. Cl₈PF shows a red shift in its absorption compared to Cl₇PF due to the electron‐withdrawing effect of the extra chlorine. No luminescence was detected for the fluorenyl radicals in CH₂Cl₂ at room temperature or 77 K, in line with TD‐DFT calculations that predict the D₁ ^–D₀ transition deep in the NIR region (1141 nm) and with very weak oscillator strength (Table S4). We note that the fluorescence at ∼680–720 nm has been recently reported for other substituted fluorenyl radicals [18], and potentially could be due to anti‐Kasha emission from the D₂ state.

TABLE 1.

Calculated and experimental photophysical data for TTM and fluorenyls.

	TD‐DFT a λabs, nm (f)	Experimental λabs, nm b	DFT a EMO, eV		Potential vs. Fc/Fc+, V b
			SOMO	SUMO	E1/2 ox	E1/2 red
TTM	371 (0.180) 467 (0.022)	373 542	−5.75	−3.44	0.81	−0.98
Cl7PF	365 (0.068) 499 (0.077) 662 (0.006)	362 546 715	−5.67	−3.78	0.70	−0.72
Cl8PF	368 (0.0548) 533 (0.077) 682 (0.003)	369 570 722	−5.77	−3.90	0.87	−0.52

^a uB3LYP/6‐31G(d), gas phase; f = oscillator strength.

^b In CH₂Cl₂.

Cyclic voltammetry (CV) of both fluorenyl radicals shows reversible oxidation and reduction waves, with the electrochemical gap (Cl₈PF, 1.39 V; Cl₇PF, 1.42 V) contracted compared to TTM (1.79 V, Figure 3d). The observed trend in electrochemical gap and the SOMO/SUMO energies is well predicted by DFT (Table 1). The lower SUMO and higher SOMO of Cl₇PF compared to TTM are explained by the greater delocalization of the frontier orbitals over the planar fluorenyl system. A ∼0.2 V positive shift of the reduction and oxidation waves in Cl₈PF compared to Cl₇PF is a result of the electron‐withdrawing effect of the additional chlorine, supporting its location in the fluorene rather than the orthogonal phenyl ring. Both Cl₇PF and Cl₈PF show an EPR signal similar to that of TTM, at a slightly higher g‐value (2.0038, Figure 3e).

To support our structural assignments of the fluorenyl radicals, we characterized their corresponding closed‐shell decomposition products. When pure Cl₇PF and Cl₈PF were left standing for a few days in solution in ambient conditions, the corresponding semiquinone products Cl₆SQ and Cl₇SQ were formed and characterized by single crystal X‐ray diffraction, NMR, and HRMS. Cl₆SQ and Cl₇SQ are presumably formed via an oxidation‐hydrolysis mechanism, as previously suggested for PTM degradation to the corresponding semiquinone (Figures 4b and S2) [42]. Notably, when the photolysis was performed in CCl₄ (the solvent used by Fox et al. [32]), these semiquinones were the major products, probably because the Cl₂ by‐product is not scavenged in this solvent, accumulates at a higher concentration, and reacts with the formed fluorenyl radicals (Figures S1 and S5). The formation of these final closed‐shell semiquinone products can also explain the discrepancies in earlier reports: the loss of EPR signal upon photolysis of PTM (previously assigned to PTM‐H formation) [31] and the marked difference between the UV‐vis absorption spectra of the photolyzed PTM solution [32] versus the expected absorption of PPF (Figures 3b and S12).

FIGURE 4.

(a) Structures of semiquinone products of photolysis with thermal ellipsoid plot from single crystal structures at 50% probability; Cl₇SQ: CCDC 2504503; Cl₆SQ: CCDC 2493348. (b) Suggested mechanism of semiquinone formation in ambient conditions.

2.3. DFT Study of the Photolysis of Other Triarylmethyls

An important remaining question is whether the calculated photocyclization activation energies can predict the experimentally observed photostabilities of various TAM radicals. To answer this question, we performed DFT calculations of TS1 for several representative species, with pyridyl, cyano, carbazole, or methoxy substituents, and compared them to the previously reported photodecomposition half‐lives (t_1/2) relative to TTM (measured under the same conditions, Table 2).

TABLE 2.

Calculated activation barriers for photocyclization and experimental photodecomposition half‐lives of representative TAM radicals.

Compound		ΔG ‡,symm, kcal/mol a	ΔG ‡,asymm, kcal/mol b	k 1/2 TTM/k 1/2 c	Experimental t1/2/tTTM 1/2 d	Experimental τ D1 (ns)
	TTM	6.3	—	1	1	5.3 (c–C6H12) [41]
	TTM‐CN2	7.7	8.2f	1.1 × 101	1.4 × 101 (PhCH3) 4.4 × 101 (CH2Cl2) [19]	49 (c–C6H12) [19]
	CM3T	8.9	n.d.	8.1 × 101	2.5 × 101 (PhCl) [41]	3.7 (c–C6H12) [41]
	PyBTM	8.0	8.9	1.8 × 101	7.1× 101 (CH2Cl2) 4.5 × 101 (c–C6H12) 8 (EtOH) [22]	6.4 (CH2Cl2) [22]
	Py3TM	14.9	—	2.0 × 106	1 × 104 (CH2Cl2) [27]	3.0 (CH2Cl2) [27]
	TTM‐Cz	12.4	11.6	7.7 × 103	5 × 101 (c–C6H12) [43] 1.5 × 103 (PhCH3) [44]	42 (c–C6H12) [43]
	TTM‐3PCz	11.8 e	10.0 e	5.2 × 102	1.0 × 104 (c–C6H12) f	21 (PhCH3) [5]
	PTM‐3PCz	9.9	10.1	4.4 × 102	1.5 × 102 (c–C6H12) [45]	26 (c–C6H12) [45]

^a Photocyclization between two identical rings.

^b Photocyclization between two different rings.

^c Calculated from the DFT activation energy (TS1‐D₁) using the Eyring equation.

^d Reported experimental photodecomposition half‐life relative to TTM in the same solvent.

^e Calculated on TTM‐3Cz (Table S5).

^f This work (Figures S15,S16).

It should be noted that the experimentally determined relative photostabilities (t_1/2 /t^TTM _1/2) are also affected by the D₁ lifetime (τ), impurities [29] and the experimental conditions, including the solvent (sometimes dramatically; Table S1). Furthermore, the overlap of the UV‐vis absorption of the fluorenyl and semiquinone products with that of TTM introduces an additional source of error in the relative half‐lives of most triarylmethyl radicals (Table S1), as many of the reported decomposition curves do not follow the expected first‐order kinetics. To minimize such errors, we recommend calculating photodecomposition half‐life times based on the initial rates (first 10%–20% degradation).

Nevertheless, the gas‐phase calculated activation energies show a reasonable correlation with the measured photodecomposition half‐lives in solution. The most photostable derivatives (TTM‐Cz, TTM‐3PCz, and Py₃TM) are predicted to have the highest activation energies (≥10 kcal/mol), suggesting that DFT calculation of the first transition state is a useful tool for the design of photostable trityl radicals.

2.4. Photolysis of TTM‐3PCz

As carbazole‐appended TTMs have become the most commonly used luminescent radicals, we synthesized and photolyzed TTM‐3PCz as a “benchmark” with high quantum yield, high photostability, and widespread applications [5, 26]. After photolysis in 0.007 M dry degassed cyclohexane under argon atmosphere, there was a complete loss of TTM‐3PCz's red fluorescence, and 3PCzPF was isolated by column chromatography at 70% yield (Figure 5). HRMS of 3PCzPF showed a single major peak corresponding to C₃₇H₁₈Cl₆N, corresponding to the loss of two chlorines by TTM‐3PCz, indicating the same photocyclization product as observed for TTM. The smaller peak at +34 Da corresponds to C₃₇H₁₇Cl₇N, the 1,8‐sigmatropic shift fluorenyl product, which could not be separated chromatographically. One could expect two possible isomers of 3PCzPF produced via symmetric and asymmetric cyclization of TTM‐3PCz (Figure 5a). DFT calculations (Table 2) suggest a preference for the formation of asymmetric 3PCzPF‐A, although the difference is relatively small and sensitive to the used functional (Table S5), so both isomers could be present in a mixture. The CV shows a pair of oxidation peaks, in ca. 1:1 ratio, at ∼0.40 and 0.62 V versus Fc/Fc⁺, as expected for the two isomers based on the DFT calculations (Figure 5c and Table S7). Compared to Cl₇PF/Cl₈PF photoproducts of TTM, 3PCzPF radicals appear to be more stable, which, together with a greater variety of possible isomers (Figure S5), prevented us from isolating and characterizing the corresponding semiquinone products. Regardless, these results indicate that photocyclization remains a major photodecomposition mechanism for even the most photostable, carbazole‐containing TAM radicals.

FIGURE 5.

(a) Possible isomers of 3PCzPF produced by photodecomposition of TTM‐3PCz with DFT‐predicted oxidation potentials (Table S7). (b) Mass spectrum of 3PCzPF, APCI negative mode. (c) Cyclic voltammetry of 3PCzPF and TTM‐3PCz (arrows indicate starting point and direction of scan). (d) X‐band EPR spectrum of TTM‐3PCz and 3PCzPF in CH₂Cl₂.

3. Conclusions

In summary, we have studied the photolysis of TTM for the first time and demonstrated that it proceeds through a 5‐electron photocyclization, followed by an unusual 1,8‐sigmatropic chloride shift affording two fluorenyl radical products. These radicals slowly decompose when exposed to air (and in anaerobic conditions, by reacting with Cl₂ by‐product) to afford closed‐shell semiquinones, whose structure was confirmed by X‐ray crystallographic analysis. This further decomposition resolves conflicting reports in the prior literature of PTM's photolysis. Experimental and computational data support the finding that other triarylmethyl radicals, including benchmark derivatives (TTM‐3PCz), undergo the same photodegradation as TTM. We propose that DFT analysis of the photocyclization transition state can guide the design of photostable triarylmethyl luminescent radicals.

Conflicts of Interest

The authors declare no conflicts of interest.

Supporting information

Supporting File 1: anie71404‐sup‐0001‐SuppMat.pdf.

Supporting File 2: anie71404‐sup‐0002‐SuppMat.zip.

Supporting File 3: anie71404‐sup‐0003‐SuppMat.zip.

Data Availability Statement

The data that support the findings of this study are available in the Supporting Information of this article.