Potassium Poly(Heptazine Imide): Transition Metal‐Free Solid‐State Triplet Sensitizer in Cascade Energy Transfer and [3+2]‐cycloadditions

Abstract Polymeric carbon nitride materials have been used in numerous light‐to‐energy conversion applications ranging from photocatalysis to optoelectronics. For a new application and modelling, we first refined the crystal structure of potassium poly(heptazine imide) (K‐PHI)—a benchmark carbon nitride material in photocatalysis—by means of X‐ray powder diffraction and transmission electron microscopy. Using the crystal structure of K‐PHI, periodic DFT calculations were performed to calculate the density‐of‐states (DOS) and localize intra band states (IBS). IBS were found to be responsible for the enhanced K‐PHI absorption in the near IR region, to serve as electron traps, and to be useful in energy transfer reactions. Once excited with visible light, carbon nitrides, in addition to the direct recombination, can also undergo singlet–triplet intersystem crossing. We utilized the K‐PHI centered triplet excited states to trigger a cascade of energy transfer reactions and, in turn, to sensitize, for example, singlet oxygen (1O2) as a starting point to synthesis up to 25 different N‐rich heterocycles.


Introduction
Artificial photosynthesis has been the primary area of inorganic semiconductors application in chemistry for many years. [1][2][3] Similarly,c arbon nitrides are traditionally associated with water splitting and CO 2 conversion. [4][5][6] In fact, the evolution of hydrogen from water in the case of polymeric carbon nitride impacted work in related areas,s uch as photoelectrochemical cells, [7] metal-free electrodes, [8,9] and electroluminescence devices. [10] Applications beyond the aforementioned are autonomous actuators, [11] photodetectors based on photon driven ion transport in asymmetric carbon nitride membranes, [12] and light-driven ion pump. [13] Polymeric carbon nitride materials have also been recognized as versatile and reliable heterogeneous photocatalysts for preparing value-added organic compounds. [14,15] Ghosh, Kçnig et al. showed, for example,that mesoporous graphitic carbon nitride (mpg-CN) enables various kinds of reactions and also enables one-pot C À Hb ifunctionalization of organic molecules. [16] Extending this first generation catalysts,p otassium poly(heptazine imide) (K-PHI), ac rystalline carbon nitride material, facilitates aremarkable number of unique reactions. Leading examples are oxidative thiolation of toluene at room temperature [17] and multiple tandem reactions. [18,19] It is also capable to store electrons. [20,21] In the context of carbon nitride photocatalysts,t he overwhelming majority of reactions is based on electron transfer. Energy transfer,o nt he other hand, generates excited-state molecules rather than charged radicals.T herefore,n ew reaction paths can be designed. [22] In terms of dioxygen, one of the greenest oxidants in organic synthesis and simplest bimodal reactant, one-electron reduction leads to the superoxide radical (O 2 C À ), while energy transfer from at riplet excited state sensitizer affords singlet oxygen ( 1 O 2 ). Thechemistry of O 2 C À and 1 O 2 could not be more different. [23] O 2 C À participates in proton abstraction, disproportionation, or nucleophilic substitution reactions, [24] while the most prominent reactions of 1 O 2 are Diels-Alder cycloaddition and formation of dioxetanes. [25][26][27] Numerous small organic compounds have been reported to sensitize 1 O 2 :I r(ppy) 3 , [28] Ru(bpy) 3 Cl 2 , [29] [Mes-Acr] + ClO 4 À , [23] riboflavin tetraacetate (RFT), [30] just to name af ew.E ase of separation, higher thermo-chemical stability, and possible application on large scale make solid-state sensitizers more attractive than homogeneous analogues. pconjugated triplet sensitizers have been atopic of research in past decades and have been quite successfully used in areas such as light energy conversion, LEDs fabrication, [31,32] photon upconversion, [33][34][35] cells imaging. [36] However, only few examples are known using such materials in photocatalysis. [37] Notable is that their use has been restricted to "model" reactions. [38] Most of these solid-state sensitizers are made out of soft polymer matrices with encapsulated platinum group metal complexes. [39] To date the modus operandi of carbon nitrides has always been linked to electron-transfer reactions,w hile energytransfer reactions in most cases have not been even considered. Replacing hazardous and toxic oxidants by simpler and more sustainable chemicals,s uch as 1 O 2 ,i sacentral topic of contemporary research. By virtue of 1 O 2 mediated reactions, which are surprisingly restricted to the synthesis of model compounds only,m any questions and challenges evolve around the function of carbon nitrides.Finally,weenvisioned bimodal photocatalysis,f eaturing ar edox mediator,w hich is also ap hotosensitizer,i ntegrated into the same material, as atool to intensify research applied in the synthesis of organic molecules.
Herein, we address an umber of questions using K-PHI ( Figure 1). We refine the crystal structure of nanocrystalline K-PHI, giving special attention to understanding the defects structure of this compound, and provided unambiguous evidence that light-excited K-PHI indeed undergoes singlettriplet intersystem crossing (ISC).
As such, we employ ac ascade of energy-transfer reactions,starting with the K-PHI triplet excited states to O 2 and as ubsequent quenching with aldoximes.I nteraction of 1 O 2 with aldoximes gives nitrile oxides,w hich triggered our interest in exploring the synthesis of various oxadiazoles-1,2,4 and isoxazoles via dipolar [3+ +2]-cycloaddition.

Results and Discussion
K-PHI was prepared from 5-aminotetrazole in LiCl/KCl eutectics using mechanochemical pre-treatment of the respective precursors, [40] while details regarding its characterization are given in Figure S1 in the Supporting Information. At first glance,t he collected X-ray powder diffraction patterns of K-PHI revealed both anisotropy in the shape of the Bragg peaks (in the 2q range 20-428 8)a nd ac learly pronounced diffuse halo,t hus pointing towards the presence of defects/disorder (Figure 2a). Having ab etter understanding of the real structure of K-PHI and the degree of structural disorder is considered essential for explaining and tuning K-PHI properties.
To start building as tructural model, ah igh-resolution transmission electron microscopy (HRTEM) study was performed. K-PHI powder consists of lamellar nanocrystallites, which form large agglomerates (Figure 3a). Fast Fourier Tr ansforms (FFTs) obtained from the HRTEM images can be indexed in ah exagonal lattice with unit cell parameters a = 11.4(8) and c = 3.7 (2) .HRTEM images reveal alayered structure with nanometer-sized domains that display unit cell  distortions,faults in the sequences of CN-layer stacking, edge and screw dislocations as well as rippling of CN-layers ( Figure 3).
TheX RD pattern of K-PHI can be correspondingly indexed in ah exagonal lattice with unit cell parameters: a = 12.637(3) , c = 3.2998(3) ,space group P31m (157), which is in good agreement with TEM data. Based on HRTEM, XRD data, and general assumptions about the structure from earlier work, [41] the starting model for refinement was proposed to consist of heptazine units,w hich are placed on top of each other in as o-called AAA stacking forming continuous channels along the c direction. Models with different CN-layers stacking, in particular ccp (ABCABC), hcp (ABAB) and mixed stacking (AABB,e tc.) allowed the positions and the broadening of the peaks to be partially described, but did not give ab etter description of the XRD pattern, suggesting the presence of stacking faults rather than as econd structural modification in the sample.P otassium atoms were located closer to the center of the channels and in between layers;their positions were corrected further during refinement based on difference Fourier maps.T he details of the refinement, refined atomic coordinates,and temperature factors can be found in Tables S1 and S2.
Our results are in good agreement with the model presented by Lotsch et al. [42] However,w eb elieve that describing the nanocrystalline sample based on the model obtained for larger crystals [42] would not be totally correct in our case.W et hink that ah igh symmetry description with defects and disorder included is closer to the real structure of the nanocrystalline sample and assists in explaining the high catalytic activity of K-PHI.
From Table S2 we derive that the highest probability of finding any Katoms corresponds to the K(1) positions (brown spheres on Figure 2), suggesting that Ka toms tend to sit closer to the center of the channels so that they can form bonds with bridging nitrogen atoms N(1).
Note that the structural model obtained is still idealized; in the real compound, there is ahigh degree of disorder, which is associated with 1) the stacking of PHI layers,a nd 2) a disordered distribution of Ka toms (the low probabilities to find K(2) and K(3) at particular position suggests that atoms can be randomly distributed in between layers). Taking earlier reports into account, an optimum size of K-PHI crystallites exists in terms of highest-performing photocatal-ysis,which is based on electron transfer, such as dehydrogenation of alcohols. [40] Diffuse reflectance UV-vis (DRUV-vis) spectrum suggests that in the K-PHI structure at least two band gaps exist ( Figure 4a). Of great relevance is the onset of absorption with an energy of 2.64 eV; it relates to the intrinsic optical band gap seen typically for graphitic carbon nitrides. [4] As mooth onset at around 1.86 eV stands for low-energy transitions, which involve intraband states (IBS). [43] Room-temperature steady-state photoluminescence (PL) spectrum of K-PHI is more complex compared to,f or example,g -C 3 N 4 , [43] and involves at least 4major transitions. Thep eaks at 2.76 eV,2 .53 eV,a nd 2.43 eV correspond to CBM-to-VBM transitions whereas that at 2.05 eV is assigned to CBM-to-IBS transitions (Figure 4b).
To correlate the 600 ps t FL , [17] which is significantly shorter than typically observed for carbon nitride materials,( normally over 1ns), with the photocatalytic activity of K-PHI (shown below), we performed further spectroscopic characterization. Thef luorescence internal quantum efficiency (IQE) of K-PHI is 0.072 %. Relatively low IQE values speak for non-radiative deactivations in carbon nitride once in the singlet excited state.T aking the extremely short fluorescence lifetime of 600 ps into account, we conclude that only 1out of 1000 exciton separation events leads to recombination and that the vast majority of excitons reaches non-radiative trap states in amatter of sub-nanosecond time frame.However,it still remains unclear what is the quantum state of those.
By applying a10msdelay,the phosphorescence spectrum was recorded (Figure 4b). Thephosphorescence features are red-shifted compared to the fluorescence.G iven that fluorescence spectrum of K-PHI comprises at least four transitions,w ith 47.7 %c ontribution of the peak at 2.05 eV to the total fluorescence (Table S3), ISC is more favorable when additional states are associated with this transition. In this view,t he singlet-triplet energy gap in K-PHI was calculated to be 0.2 eV (Figure 4b). Earlier reported singlet-triplet energy gaps for graphitic carbon nitride range from 0.156 to 0.248 eV. [38] To address the question on the quantum state of excitons in non-radiative traps and to obtain insights into the excited state dynamics in K-PHI, we conducted transient absorption measurements (Figure 4c and Figure S6). Upon excitation at 387 nm, abroad minimum centered at 525 nm and amaximum at l > 1100 nm evolve.T he minimum is assigned to ground state bleaching of the singlet excited state,w hile the maximum is assigned to excited-state absorptions.W ithin afew hundred picoseconds,all of the aforementioned features transform into abroad negative transient stretching over the entire spectrum and minimizing at around 600 nm. This signal decays to zero after 10 ms. Thes teady deactivation of the short-lived transient together with the evolution of the longlived transient points to ap opulation of at riplet from the former singlet excited state.
Multi-wavelength analysis of the time-absorption profiles yields five lifetimes (Figure 4d). Theg round-state recovery from the singlet excited state occurs with two lifetime components of t 1 = 1.4 ps and t 2 = 12.1 ps.Atriplet excited state is populated via intersystem crossing within alifetime of t 3 = 470 ps.T he triplet finally deactivates with two lifetime components t 4 = 161 ns and t 5 = 1.9 ms. Thep resence of two lifetime components indicates the presence IBS.
Such ac omplex excited-state dynamics of K-PHI is very different from ac arbon nitride reported by Wu et al. and Durrant et al.,w ho described the excited state deactivation by as ingle exponential fitting function. [44,43] On the other hand, Xie et al. also observed by TASf ormation of triplet states in carbon nitride,a lbeit intersystem crossing occurs within afew picoseconds. [45] Figure 4esummarizes the excited state dynamics in K-PHI.
To explore the potential of the non-radiative triplet excited states in K-PHI to sensitize,f or example, 1 O 2 ,w e monitored the emission of K-PHI suspension in MeCN saturated with O 2 at 1271 nm. Indeed, we recorded the distinct 1 O 2 fluorescence pattern (Figure 4f). mpg-CN was also able to produce 1 O 2 ,a lbeit with significantly lower efficiency.S ensitization of 1 O 2 thereby evidences that triplet states in carbon nitride materials may be used for energy transfer in photocatalytic reactions as will be shown below.
To characterize the non-radiative trap states,w einvoked theoretical modelling of the K-PHI structure.T od ate,t he density functional theory (DFT) modeling has been carried out for diverse carbon nitrides. [43,46] Considering,however, the unique chemical structure of K-PHI, we want to obtain acohesive picture of its electronic structure.
Using the refined crystal structure,weperformed periodic DFT calculations to obtain the total and partial density of states (DOS) of K-PHI. Forour calculations,wetook the K-PHI structure bearing three K(1) potassium atoms (those with af raction 0.7032 in Table S2, brown spheres on Figure 2b). TheDOS profile (Figure 5a)reveal that the conduction band of K-PHI is formed by s-and p-orbitals of nitrogen (55 %) and carbon (26 %) followed by potassium s, p, and d-orbitals (19 %) with the edge located at À0.81 eV.The experimentally determined flat-band potential is À0.50 Vversus the normalized hydrogen electrode (NHE). [47] By means of ultraviolet photoelectron spectroscopy (UPS), the valence band maximum is determined at + 2.68 eV (Figure 5b). However,s ignal starts to develop at around + 1.45 eV.T his is indicative for the existence of IBS. Taking UPS and DRUV-vis data as well as theoretical DOS modeling of K-PHI (Figure 5a)i nto account, we conclude that the IBS are located at approximately + 1.45 eV.I nt he DOS of K-PHI, IBS are observed as an "island" emerging at + 1.00 eV and stretching up to + 2.42 eV (Figure 5a,c). Presence of a" valley" between + 2.42 eV and + 2.63 eV underlines that the states are well-separated from the VB.The IBS are not caused by the introduction of potassium itself,as their corresponding contributions in the range of energies from + 0.93 to + 2.5 eV are negligibly small (0.48 %). Instead, we ascribe them to n-p*t ransitions. [48] In K-PHI, the heptazine units are arranged in large macrocycles and each of them is formed by 6heptazine units (Figure 2b). Therefore, the K-PHI structure is more flexible compared to the conventional graphitic carbon nitride,i nw hich heptazine units either form ar igid conjugated 2D structure or are best described by chains of heptazine units bound by hydrogen bonds. [49] Therefore,e nsembles of heptazine units in K-PHI can adopt out-of-plane conformations,w hich, in turn, facilitates the otherwise restricted n-p*transitions. [50] Figure 5c summarizes the experimentally determined CBM and VBM in K-PHI with that obtained from the DFT modelling. Thed ifference between CBM (À0.50 eV calculated taking into account Mott-Schottky plots) and VBM (+ 2.68 eV,determined from UPS) is 3.18 eV-the energy gap or transport gap,islarger than the optical band of 2.64 eV. [51] Thep reliminary results of 1 O 2 addition to 9,10-diphenylanthracene to undergo endoperoxide formation point to higher selectivity of the heterogeneous photocatalysis versus homogeneous system (Table S4). To test the K-PHI-assisted 1 O 2 sensitization in the synthesis of N-rich heterocycles,w e examine the synthesis of oxadiazoles-1,2,4 in greater detail owing to the importance of this class of organic compounds for medicinal chemistry and material science. [52,53] We expect formation of nitrile oxides from the corresponding aldoximes upon 1 O 2 quenching followed by [3+ +2] cycloaddition to the C N-group of an itrile.
Thorough reaction-condition screening is given in Tables S5-S12. Aseries of oxadiazoles-1,2,4 (1-23, Figure 6) was prepared by coupling different oximes with nitriles.G enerally,the yields of oxadiazoles were higher combining aromatic nitriles with electron deficient oximes ( Figure S10). For ethylcyanoacetate,c yclization toward oxadiazoles 18 and 19 competes with the Knoevenagel condensation. Using the developed method, we prepared several applied oxadiazoles derivatives. [53,54] An attempt to couple oxime of 3-formylbenzoic acid with 2-fluorobenzonitrile in order to prepare PTC124-a pharmaceutical drug to target genetic disorderswas not successful. [55] Instead, we succeeded in synthesizing Figure 5. Band structure of K-PHI. a) Partial (C, N, K) and total DOS in K-PHI ground state. Fermi energy is located at 0eV. b) UPS of K-PHI. c) Experimental band structure of K-PHI and that derived from DFT calculations. TDOS on the right is given as ag uide for CBM, IBS, and VBM. Experimentald ata set:CBpotential determined by Mott-Schottky analysis. [47] Uncertainty of IBS onset and VBM determination due to smooth signal onset in UPS and presence of several emissive states near CBM (Figure 4b)are denoted by closely lying levels. [a] Blue phosphorescent OLEDs precursor. [53] [b] Ap otential drug for treatment of Alzheimer's disease. [54] [c] A pharmaceutical drug (PTC124) precursor to target genetic disorders. [55] [d] NMR yield. the corresponding methyl ester 22.T he ester group of the oxadiazole is subjected to hydrolysis under alkaline conditions. [56] Furthermore,w ef ound that the intramolecular cyclization in chalcone oximes is triggered by K-PHI under blue light irradiation. Thetentative intermediate,namely 4,5dihydroisoxasole,isf urther oxidized to isoxazole 24.
Ther eaction between benzaldehyde oxime and acetonitrile has been chosen to investigate the mechanism. The apparent quantum yield (AQY) of oxadiazole 11 reached (5.8 AE 0.4) 10 À5 %a fter 6hof irradiation ( Figure S11). The redox potentials of the reagents were determined by cyclic voltammetry (CV) in MeCN (Figure 7a). Acetonitrile is stable against oxidation and reduction in the range from À2.1 Vto+ 1.9 Vversus Fc/Fc + (Fc = [(h-C 5 H 5 ) 2 Fe]. Relative to the aforementioned, the standard redox potential of the O 2 / O 2 C À couple is centered at À1.27 V.
Given that the potential of the conduction band of K-PHI is located at À0.9 Vvs. Fc/Fc + (À0.5 Vvs. NHE), reduction of O 2 to afford O 2 C À is thermodynamically challenging.T herefore,e lectrochemical measurements also suggest that 1 O 2 sensitization is the preferential path of O 2 activation. Depending on the oxime structure,the onset of irreversible oxidation occurs in the range from + 0.51 V, for pyrrole-2-carbaldehyde oxime 28 a,t o+ 1.67 Vv ersus Fc/Fc + ,f or pivalaldehyde oxime 9a ( Figure S12). Reaction in the presence of electron scavengers,s uch as nitrobenzene and S 8 , [17] excluded participation of the photogenerated holes in the synthesis of oxadiazoles-1,2,4 ( Table S6). Addition of DMPO to the reaction mixture did not lead to the formation of the DMPO-O 2 C À adduct as evidenced by EPR spectroscopy ( Figure S7). On the contrary,s ynthesis of 9,10-diphenylantracene endoperoxide (Table S4), 1 O 2 fluorescence detection (Figure 4f), and detection of TEMPO radical (Figure 7b) formed in situ from 2,2,6,6-tretramethylpiperidine univocally confirmed participation of 1 O 2 in the reaction. [57] When present in equal amount, benzaldehyde oxime 2a is converted into oxadiazole 11 2.91 AE 0.15 faster than 2a-d 1 (primary kinetic isotope effect, KIE, Figure 7c). Ther esults suggest that breaking of the CÀDb ond is the rate limiting step.W hen benzaldehyde oxime 2a was mixed with CH 3 CN and CD 3 CN in a1:1 ratio,under the photocatalytic conditions using K-PHI, CH 3 -substituted oxadiazole 11 forms 1.11 AE 0.04 times faster (secondary KIE) compared to the CD 3 -substituted 17 oxadiazole.I nt he case of Ir(ppy) 3 ,t he secondary KIE is 1.31 AE 0.05. Compared to the homogeneous catalysis, K-PHI reduces the isotope effect due to the polarization of the C À Db onds induced by the electrostatic field on the K-PHI surface.Zeta potential measurements resulted in avalue of À40 mV for K-PHI. [58] Even though the vast majority of oximes quench 1 O 2 by means of dehydrogenation, anumber of electron-rich oximes including 26 a-31 a fail to undergo click-heterocyclization ( Figure S13). Instead, aldehydes are formed as the products of C = Nb ond cleavage.Alikely rationale is based on the formation of an intermediate charge-transfer complex between the electron-rich p-conjugated system (27 a-30 a)a nd 1 O 2 . [59,60] Once formed, the enriched electron density on dioxygen facilitates nucleophilic addition to the C=Nb ond followed by recovery of the aldehyde.N otably,carboxylic or phenolic protons as part of oximes 26 a and 31 a inhibit the pathway A. [61] Taking the experimental data into account, we gather at entative mechanism of 1 O 2 quenching by oximes,w hich is summarized in Figure 7d.D epending on the chemical structure of oxime,i ts interaction with 1 O 2 is likely to follow two different pathways,n amely oxidation to the nitrile oxide (pathway A) or addition to the C=Nb ond (pathway B). Pathway Ai so perative in most of the studied cases. Pathway Br equires electron-rich oximes or oximes bearing acidic protons.I ndependent support for our mechanistic conclusions came from solvent-screening tests (Table S10). In the absence of,f or example,c lick-cyclizable nitriles,p athway Bfor benzaldehyde oxime 2a is activated exclusively.In stark contrast, solvents,w hich are capable of effectively quenching 1 O 2 ,s uch as 1,4-dioxane,a nisole,D MSO,D MF, inhibit both pathways.I nt he case of 1,4-dioxane,t he conversion of oxime 2a is 46 %n ext to 19 %o f3 ,5diphenyl-1,2,4-oxadiazole,w hich evolves as the product of nitrile oxide dimerization.

Conclusion
Our investigations started with determining the crystal structure of K-PHI. We showed that it crystallizes in ah exagonal lattice with unit cell parameters: a = 12.637-(3) , c = 3.2998(3) ,s pace group P31m. We described stacking disorder in the PHI layers and the non-uniform distribution of K + -ions in the structure.S till, with highest probability of Ka toms to be found in the center of the channels so that they can form bonds with bridging nitrogen atoms N(1). Once the crystal structure of K-PHI was defined, we performed periodic DFT modelling to calculate,t he electronic band structure and the density of states.W e reached as ound agreement with the experiments in terms of intraband electronic states at around + 1.45 Vasthe origin of sizeable absorption in the near-IR enabling energy-transfer reactions.O nce excited with visible light, ab road range of photochemical responses sets in. Most interestingly is the fact that K-PHI undergoes significant singlet-triplet intersystem crossing across a0 .20 eV energy gap affording ar easonably long-lived triplet excited state.I nt erms of dynamics,t he combination of TASand TCSPC enabled deriving 600 ps for the radiative decay of the singlet excited state,470 ps for the non-radiative intersystem crossing, 2.07 msf or the nonradiative decay of the triplet excited state.
Our work was rounded off by utilizing the triplet excited states in K-PHI to sensitize 1 O 2 and, in turn, to employ 1 O 2 in its reaction with ab road range of oximes.H ereby,t wo pathways of 1 O 2 quenching exist:f irst (pathway A), it is the oxime dehydrogenation, which affords nitrile oxide and, second (pathway B), it is the addition to oxime C = Nb ond, which results in aldehyde formation. When 1 O 2 quenching is performed via pathway Ai n, for example,t he presence of reactive multiple bonds (CN, C=C), the nitrile oxides undergo [3+ +2]-cycloaddition. In total, 25 examples of oxadiazoles-1,2,4 and isoxazoles were synthesized in 11-82 % yield.
In the context of this work, the developed photocatalytic [3+ +2]-cycloaddition reaction is applicable for functionalization of polymers,such as those with pendant multiple bonds. At the same time,m olecules,o ther than O 2 ,c an be used as energy acceptors to broaden classes of organic compounds accessible via carbon nitride photocatalysis.T hese are ongoing projects.