Photocatalytic CO2 reduction with a quantum e ciency exceeding 60%: time-resolved spectroscopic and X-ray studies on Cu(I) photosensitizers in coordinative interaction with Co(II) phthalocyanine catalysts

Jia-Wei Wang Sun Yat-Sen University https://orcid.org/0000-0003-1966-7131 Xian Zhang University of Göttingen Michael Karnahl Technische Universität Braunschweig Zhi-Mei Luo Institute of Chemical Research of Catalonia Zizi Li Sun Yat-sen University Yanjun Huang Sun Yat-sen University Jin Yu Argonne National Laboratory Wenhui Hu Marquette University Xiaoyi Zhang Argonne National Laboratory https://orcid.org/0000-0001-9732-1449 Dooshaye Moonshiram IMDEA Gangfeng Ouyang (  cesoygf@mail.sysu.edu.cn ) Sun Yat-sen University https://orcid.org/0000-0002-0797-6036


Introduction
Sunlight-driven reduction of CO2 continues to attract attention as a sustainable, carbon-neutral route to produce renewable fuels. 1,2 Due to the sluggish multi-electron reaction kinetics and strong competition with H2 evolution, 3,4 considerable efforts have been dedicated to the development of efficient and selective systems for photocatalytic CO2 reduction. Most examples, fully or in part, utilize metal complexes as molecular photosensitizers (PSs) or catalysts, which are highly optimizable via synthetic methods and easy to characterize for mechanistic investigations. 5-7 Among them, noble-metalbased PSs, like Ru, Ir and Re complexes, are extensively applied to drive the catalysts. [8][9][10][11] However, the high expense and low abundance of 4d and 5d transition metals inevitably precludes large-scale applications of such systems and therefore earth-abundant molecular systems are of intensified interest. [12][13][14][15][16] Organic dyes could serve as economical PSs due to their strong absorption and high redox potentials. An early example was contributed by Fujita et al., 17 in which the photocatalytic reduction of CO2 to CO and HCOOH using Co macrocycles as catalysts was assisted by p-terphenyl, giving a quantum efficiency (Φ) of 25% at 313 nm. More recently, Robert and co-workers have exploited several organic dyes like 9-cyanoanthracene, 18 purpurin 19 or 3,7-di(4-biphenyl)-1-naphthalene-10phenoxazine 20 in combination with iron porphyrin catalysts. These examples demonstrate the high catalytic efficiency and selectivity of organic dyes, while they may suffer from low photostability, 21 demanding new alternatives for CO2 reduction.
For instance, Cu(I) based PSs have shown great potential in photocatalysis caused by their versatile redox characteristics and highly variable ligand scaffolds for tunable redox and excited-state properties. [21][22][23][24][25][26] In terms of CO2 reduction, Ishitani et al. have designed a diphosphine-tethered phenanthroline ligand to prepare a dimeric Cu(I) PS, which displayed notable stability during photocatalysis along with Fe 27,28 or Mn 29 catalysts. The tetradentate nature of the applied ligand effectively prevents the dissociation of diphosphine donors around the Cu centers, achieving a maximum Φ of 57% for the co-production of CO and HCOOH. Beller and co-workers deployed the Cu(I) dyes in situ by directly mixing the Cu(I) precursor and respective ligands in the photocatalytic systems for CO2 reduction, in conjunction with cyclopentadienone Fe complexes 14 or Mn diamine complexes 30 respectively as catalysts. Furthermore, Sakai et al. 13,31 prepared a sulphonated Cu(I) PS to work in fully aqueous solutions with water-soluble Co porphyrins, achieving good selectivity (near 90%) and turnover numbers (TONs, up to 4000).
The examples mentioned above have shown some promising catalytic performances and demonstrated that Cu(I) PS are suitable for light-driven CO2 reduction. Nevertheless, there is still substantial room for improving the performances, as the noble-metal containing systems have demonstrated TON > 10 7 32 and Φ > 80%. 33 The anticipated improvement can be achieved by rationally optimizing the catalysts and PSs and additionally by enhancing the essential electron transfer between PS and catalyst. Focusing on the latter issue, various strategies have been explored, including the covalent attachment, [34][35][36] or H-bond interactions, 37 in addition to coordinative 38 interaction developed by us. However, most of these systems still rely on noble-metal-based PSs or catalysts. In contrast, fully noble-metal-free molecular systems applying weak interactions between PS and catalyst to improve electron transfer, to the best of our knowledge, have seldom been reported so far.
Herein, we have chosen heteroleptic diphosphine-diimine Cu(I) complexes, [Cu(P^P)(N^N)] + , as PSs, and cobalt phthalocyanine (CoPc) derived complexes as catalysts, for optimizing photocatalytic CO2 reduction. The three key factors, i.e. catalyst, PS and their interaction, are all considered for significant improvements in photocatalysis. In consequence, the optimal combination of [Cu I (xantphos)(bcp)]PF6 (CuBCP; xantphos = 9,9-dimethyl-4,5-bis(diphenylphosphino)xanthene, bcp = bathocuproine) and cobalt tetracarboxylphthalocyanine (CoTCPc) enables an extremely high maximum Φ of 63.5% at 425 nm for the CO2-to-CO conversion with 99% selectivity. Further, the coordinative interaction between Cu(I) PS and CoTCPc has been constructed by using a pyridine-decorated 1,10-phenanthroline-derived ligand for the Cu(I) centers, which nonetheless led to a decreased efficiency. As evidenced through DFT calculations and time-resolved X-ray absorption spectroscopic measurements, the increased flattening of the overall excited state geometry and the larger torsional angle change of the photoexcited state impair the PS stability significantly in this case. This methodical study conducted is promising for future rational developments of both earth-abundant PSs and catalytic complexes for CO2 reduction with additional interactions.

Electronic structures of Co(II) catalysts
First, a systematic series of different Co(II) catalysts (see Figure 1) with different substituents for electronic modulation were selected, namely cobalt β-tetrasulphophthalocyanine (CoTSPc), CoTCPc, CoPc, cobalt β-tetraaminophthalocyanine (CoTAPc) and cobalt β-tetra(dimethylamino)phthalocyanine  Table 1) and UV-vis spectroscopy (Supplementary Figure 4 in Supporting Information). As the structures of these β-substituted CoPc derivatives cannot be defined by single-crystal X-ray diffraction due to the different regioisomers arising from the synthesis with two random β-positions, DFT calculations were applied to analyze the structural and electronic effects of the different substituents by fixing them at the positions of 1, 5, 9 and 13. As can be seen from Figure 1 and Table 1, the β-substituents result in negligible distortions on the phthalocyanine rings with the nearly 0° angles of plane distortion φ. According to the Hirshfeld population 39 listed in Table 1, the Co atomic charges in these molecules vary through incorporation of different substituents. All Co atomic charges are positive, which indicates that the electron density is transferred from cobalt to the phthalocyanine substituents. It is found that the Co atomic charges in the complexes with the -N(CH3)2, -NH2, -H, -COOH, and -HSO3 substituents are 0.1959, 0.1971, 0.2038, 0.2084, and 0.2122 e, respectively. This is in the exact order of the increasing ability of the β-substituents to attract electrons (-N(CH3)2 < -NH2 < -H < -COOH < -HSO3) and demonstrates the facile modulations of the electronic properties of these Co(II) catalysts.

Structural and photophysical properties of Cu(I) PSs
Subsequently, four heteroleptic Cu(I) complexes of the type [Cu(P^P)(N^N)] + were prepared as PSs    (Table 4). The triexponential decay traces of the excited CuPBCP and CuPBCP' are consistent with markedly short lifetimes < 100 ns, which suggest the presence of several decay pathways and will be further discussed later. To sum up, the Cu(I) PSs bearing the xantphos or bcp ligands display higher Φem values and excited state lifetimes than those with DPEphos or pbcp, respectively.
Furthermore, cyclic voltammetry in combination with the spectroscopic results discussed above was applied to evaluate the excited-state redox potentials 46

Modulation on catalysts for photocatalytic CO 2 reduction
With basic understanding of the above Cu(I) PSs and Co(II) catalysts, they were then used for the visible-light-driven reduction of CO2 using a mixture CH3CN/TEA and BIH (1,3-dimethyl-2-phenyl- was added to the catalytic mixture to inhibit its dissociation during photocatalysis. We first screened the Co(II) catalysts together with CuBCP as the PS (Table 2, entries 1-5) under 450 nm irradiation. All five Co(II) catalysts enabled the selective formation of CO with only trace amounts of H2 and no liquid products, providing CO selectivity over 97%. CoTCPc showed the highest catalytic performance in terms of CO yield, followed by CoTSPc, CoPc, CoTAPc and CoTDMAPc ( Figure 3a). As a result, an impressive increase of more than ten-fold in catalytic efficiency is achieved from CoTDMAPc to CoTCPc. Hence, the usage of electron-withdrawing β-substituents tend to increase the catalytic performances of Co(II) catalysts, with CoTCPc as the optimal one, while the more electronwithdrawing CoTSPc shows a lower activity.
It should be noted that all components are necessary to afford a substantial amount of CO in the representative CuBCP/CoTCPc system (Supplementary Table 5), and some factors were further considered for a systematic optimization of the photocatalytic system applying the best catalyst, i.e.

CoTCPc.
As depicted by Figure 3b and S15, a decreasing concentration of CoTCPc can significantly increase the TON with a maximum of 2950 at 0.2 μM within 1 h (Table 3, Figure 18), evidencing that the CO was produced from the reduction of CO2 rather than from the decomposition of organic components. Overall, the above photocatalytic results demonstrate the extraordinary performance of the CuBCP/CoTCPc system for photocatalytic CO2-to-CO conversion.

Modulations on PSs and interaction for photocatalytic CO 2 reduction
Next, the four different Cu(I) PSs were tested with the best CoTCPc catalyst under the same optimized catalytic conditions (Figure 3e and Table 2, entries 5, 7-9). The activity was found to follow a descending order of CuBCP > CuBCP' > CuPBCP > CuPBCP'. It is obvious, that the two Cu(I) PSs bearing the xantphos ligand exhibit higher performances than their counterparts with DPEphos.
The differences can be ascribed to the bulkier and more rigid nature of xantphos than DPEphos. The bulky xantphos ligand can protect the copper center against undesired exciplex quenching by solvent molecules or counter ions. Moreover, xantphos is less flexible, which prevents the unwanted flattening in the excited state. Accordingly, it has been reported that in terms of photocatalytic hydrogen 11 evolution, the replacement of the DPEphos ligand by the xantphos not only enhanced the (photo)stability, but also improved the photophysical properties (e.g. excited-state lifetimes). 24,59 Here we also observed that the photophysical properties of CuBCP, including Φem and lifetimes, are better than those of CuBCP', which are consistent with the observed activity order, demonstrating that the photophysical properties of Cu(I) PSs play a key role in CO2 photoreduction.
The most notable observation is that the use of the pyridine-containing pbcp ligand does not result in an improved reduction of CO2, which is most possibly caused by the much shorter excited state lifetimes of CuPBCP and CuPBCP'. Nevertheless, our previous studies indicate that a slightly inferior PS can be compensated by reinforcing the intermolecular electron transfer via coordinative interaction. 38 As a type of covalent bonds, the labile coordinate bond not only endows the dynamic stability in the coordinatively interacted system, but also additionally facilitates the electron transfer processes in a manner of inner-sphere electron transfer (ISET) via the coordinative interaction 60 as either a stable adduct or a transition state, potentially circumventing the diffusion limit. 61 Therefore, we also examined the coordinative interaction by 1 H NMR titration experiments. Prior to titration, the proton assignment for CuPBCP was carried out in d7-DMF (Supplementary Table 2 Table   7). This value is in the same order of magnitude and in the same binding model as the TEA-free system, demonstrating the tolerance of coordinative interaction against competitive TEA coordination.
According to the above results, substantial coordinative interaction has been achieved between CuPBCP and CoTCPc, but this cannot compensate for the impaired photophysical properties caused by the replacement of bcp with pbcp, leading to diminished activity. The further reasons for the loss of activity caused only by the replacement of two C atoms by N atoms should be investigated in detail for the rational design of Cu(I) PSs, which will be discussed later.

Structure-activity relationship
Aforementioned electrocheical results have illustrated that the electron-withdrawing effects effectively impose positive shifts of Co II/I redox couple, which is amenable for photocatalysis at lower overpotentials. 54 However, CoTSPc, with the most electron-withdrawing sulfonic-acid substituents, gave a minor performance than CoTCPc. with the order of catalytic activity. In contrast, the order of the ΔG(*CO) values is more consistent with the order of catalytic activity (Figure 4), which suggests that the CO desorption step plays a crucial role for the catalytic performance. The key contribution of CO desorption ability in CO2 reduction has been extensively documented. 66,67 This DFT-calculated volcano-type plot computationally indicates CoTCPc as the best catalyst in our case, while this primary observation demands more CoPc derivatives as well as more detailed theoretical/experimental works, which are on the progress, for further confirmation.

Photo-induced electron transfer
The proposed catalytic mechanism and the observed structure-activity relationship were further supported by verifying the underlying photo-induced electron transfer processes. For this purpose, we focused on the investigation of the excited states of CuBCP and CuPBCP to also analyze the low performances caused by the pyridine pendants. Firstly, quenching experiments with the excited state of the Cu(I) PSs were carried out by steady-state fluorescent spectroscopy ( Figure 5 and Table 3). On one hand, both CuBCP and CuPBCP can be effectively quenched by BIH (Figure 5a and 5b Cu PS − + Co II TCPc → Cu PS + Co I TCPc (3) Cu PS * + Co II TCPc , �⎯� Cu PS + + Co I TCPc (5) Cu PS + + BIH → Cu PS + BIH + (6) We then implemented transient absorption (TA; Figure 6 and Table 3     fitting for the CuPBCP system with BIH. The data were collected by following the spectra at 360 nm in Ar-saturated CH 3 CN upon excitation at 420 nm.

Photocatalytic CO 2 reduction with other molecular catalysts
Besides the catalytic properties and mechanisms, the above systematic studies also reveal that the   (Table 4 and  Interestingly, clear changes were not only observed in the oxidation state of Cu but also in its electronic configuration and coordination geometry. MLCT transition causes a 3d electron to be promoted to the low-lying π* orbital of the ligand upon excitation, thus leading to a d level 19 occupational change from 3d 10 to 3d 9 . A pre-edge feature corresponding to the 1s to 3d quadrupole transition which gains intensity due to 4p mixing 78,79 into the Cu 3d shell is thus observed at low photon energies at 8979 eV (Figure 7).   The kinetics of the decay of the 3 MLCT excited state were additionally monitored by fixing the Xray photon energy at 9000 eV and varying the time delay between the laser and X-ray pulses ( Figure 7 inset and Supplementary Figure 29). The 3 MLCT states of CuBCP and CuPBCP form promptly in less than the ~100 ps pulse duration of the X-rays and decay within 287 ± 63 and 64.3 ± 22.1 ns, respectively, in agreement with the previous time-resolved fluorescent experiments (Supplementary   Table 4 and Supplementary Figure 13). The excited state fractions of CuBCP and CuPBCP were 21 determined by comparing their laser-off and laser-on XANES spectra (Figure 8a and 8c) to those of previous reported heteroleptic Cu complexes 80 with similar coordination environments. A relative chemical shift in energy of 2 eV was typically observed between Cu(I) and Cu(II) reference complexes, such that the proportions of excited state of 35% and 27% were estimated in the laser on spectra of CuBCP and CuPBCP, respectively, and used to plot the actual or reconstructed XANES spectra of the excited states (Figure 8a and 8c).
The formation of a flattened geometry within the 3 MLCT state has been well known to enable nucleophilic attack by solvent molecules or counterions thus promoting the formation of a low-energy excited state known as an "exciplex", which subsequently decays to the ground state. 70,71 Alternatively, the nature of the solvent can also modify triplet lifetimes without a direct coordination to the metal centre by tuning charge transfer excited state energies. 81 The shortened excited-state lifetime of CuPBCP versus CuBCP reflects the effect of the pyridine decoration instead of benzene. That is, the pendant pyridine rings in CuPBCP seem to account for the decreased stability and lifetime of its 3 MLCT state, as they can favour a stronger flattening of the overall excited-state geometry and a larger torsional angle change, as evidenced through DFT geometry optimization calculations and experimental EXAFS fits (Table 4 and Supplementary Table 11 Table 12). The larger torsions in the pyridine rings of CuPBCP are consequently presumed to lead to the decreased steric hindrance and enhanced solvent accessibility in its excitedstate conformation, namely a stronger exciplex effect.

Discussion
In summary, triple modulations, in terms of catalyst, PS and additional coordinative interactions, have been employed in this work to accomplish high-performance photocatalytic CO2 reduction to CO, favors the luminescent properties and thus the photocatalytic activity. On the other hand, the results of the time-resolved X-ray absorption spectroscopy and the DFT calculations indicate for the diimine N^N ligand that the replacement of phenyl by pyridyl groups enhances exciplex quenching and thus induces short-lived triplet excited states. As a proof-of-concept, the intended coordinative interaction between the pendant pyridine rings in the Cu(I) PSs and the axial sites of CoTCPc has been established, which can facilitate the electron transfer between the above two components. Interestingly, the presence of such an interaction does not lead to an increase in catalytic activity, because the performance is still limited by the unstable excited state of the pyridine-decorated PS, noting the impact of delicate ligand modifications. Based on the above merits, we believe that our work opens new avenues to the development of highly cost-efficient molecular systems for CO2 photoreduction.
The above-described experiments on the optimization of photocatalytic systems can serve as an inspiring and systematic guide to improve catalytic performances.
electrochemical workstation (CHI 620E). Gas chromatographic analysis was conducted on an Agilent 7820A gas chromatography. The isotopic labeling experiment was conducted under 13 CO2 atmosphere and the gas in the headspace was analyzed by a quantitative mass spectrometer attached Agilent