Efficient CO2 photoreduction enabled by the energy transfer pathway in metal‐organic framework

Many studies in metal‐organic frameworks (MOFs) aiming for high photocatalytic activity resort to self‐assembling both energy donor and acceptor building units in skeleton to achieve effective energy transfer, which, however, usually needs tedious synthetic procedure and design of a new MOF. In this work, we demonstrated that building a Förster resonance energy transfer (FRET) pathway can be realized through suitable molecular doping in a given MOF structure without altering the original porous structure, presenting an alternative strategy to design efficient photocatalysts for CO2 reduction. In situ electron spin resonance, ultrafast transient absorption spectroscopy, and computational studies reveal that the FRET‐induced excitation has dramatically altered the exciton transfer pathway in structure and facilitated electron‐hole separation. As a result, the molecular doped MOFs synthesized through one‐pot reaction show outstanding selectivity (96%) and activity (1314 μmol·g−1·h−1) for CO production versus almost no activity for the pristine MOFs, and this result stands out from existing competitors. Furthermore, the reaction mechanism was proposed and the intermediate signals were detected by in situ diffuse reflectance infrared Fourier transform spectroscopies. This study presents a clear picture of building FRET process in MOFs through molecular doping and provides a new design strategy for MOF‐based photocatalysts.


INTRODUCTION
9][10][11][12][13] However, reactions with CO 2 as a reactant are usually thermodynamically and kinetically unfavorable, therefore exploration of efficient photocatalysts for CO 2 reduction is of a great challenge.[16] Most strikingly, the tunable metallic and organic constituents endow MOFs with tremendous potential for optimizing the catalyst performance from both thermodynamic and kinetic aspects, and the related investigation would provide insight into rational design of effective photocatalysts.
[23][24] Hence, regulating the electronic band structure of catalysts to facilitate electron transfer in these robust frameworks poses a problematic issue in this field. [25,26]So far, designing new organic linkers, introducing co-catalysts, or combining with other active materials by post-synthesis are the commonly used strategies to improve the electron transfer and separation efficiency.However, these strategies usually need tedious synthetic procedure and show low efficiency.[29][30][31] Inspired by the natural biological process, we aim to build energy transfer pathways in MOF structures through molecular doping to boost photocatalytic activity.
Bearing these considerations in mind, NU-1000 based on tetratopic 4,4′,4′′,4′′′-(pyrene-1,3,6,8-tetrayl)tetrabenzoate (H 4 TBAPy) linkers and Zr-oxo clusters was chosen because of its excellent chemical stability, high CO 2 uptake, and the extended π conjugation.The unsaturated metal sites (8-connected) and hierarchical micro-mesoporous structure (1.3 and 3.4 nm pore sizes) provide adequate binding sites and void space for molecular doping. [22,32,33]Addition of [5,10,15,20-tetrakis(4-carboxyphenyl)porphyrinato]-Co(II) (TCPP(Co)) molecules in the NU-1000 structure (denoted as TCPP(Co)⊂NU-1000) led to a higher production rate (1314 μmol⋅g −1 ⋅h −1 ) and selectivity (96%) in the conversion of CO 2 to CO compared to pure NU-1000 and other control groups.It is noteworthy that such a boosted catalytic activity cannot achieve if the doped material was obtained through a post-synthetic strategy as usually adopted by previous studies.In addition, the catalytic reaction can proceed in aqueous media without any additives, such as photosensitizers and/or co-catalysts.Significantly, experimental results and theoretical calculations indicate that the enhanced photocatalytic performance of TCPP(Co)⊂NU-1000 can be attributed to the increased production of photoexcited electrons and efficient electron-hole separation provided by the energy transfer pathway in TCPP(Co) doped in NU-1000.

RESULTS AND DISCUSSIONS
The doped material TCPP(Co)⊂NU-1000 was synthesized via one-pot solvothermal reaction of ZrCl 4 , and H 4 TBAPy in the presence of a small amount of TCPP(Co) (Section 3 in SI).The powder X-ray diffraction (PXRD) pattern of TCPP(Co)⊂NU-1000 is consistent with that of pristine NU-1000 (Figure S1 and S2).Moreover, the N 2 and CO 2 isotherms of TCPP(Co)⊂NU-1000 are similar with those of pristine NU-1000 (Figure S8, S9 and S10), indicating that the structure and porosity is maintained after doping with TCPP(Co).Inductively coupled plasma-optical emission spectroscopy (ICP-OES) shows the Co content in NU-1000 is about 0.286 wt%, which is equal to one TCPP(Co) in every five Zr 6 clusters (see Equation S1 for detailed discussion).
The successful doping of TCPP(Co) in NU-1000 is further supported by the transmission electron microscope coupled with energy-dispersive X-ray spectroscopy, X-ray photoemission spectra, and thermogravimetric analysis (Figure S14, S15, S16, and S17).In situ diffuse reflectance infrared Fourier transform spectroscopies (DRIFTS) show a sharp peak at 3674 cm −1 derived from the -OH stretch of the Zr 6 nodes and which decreases apparently after doping TCPP(Co) in NU-1000, verifying that some -OH groups on the Zr 6 nodes have been exchanged by TCPP(Co) molecules. [34]Moreover, TCPP(Co)⊂NU-1000 also exhibited a new strong band at 1663 cm −1 which was assigned to uncoordinated -COOH group from TCPP(Co) molecules (Figure 1B).Therefore, it appears that not all the benzoate groups of TCPP(Co) molecules were bound to the framework.The ultraviolet-visible (UV-Vis) diffuse reflectance spectrum of TCPP(Co)⊂NU-1000 (Figure 1C) shows a new absorption feature in the long wavelength region near 550 nm, and the band gap of TCPP(Co)⊂NU-1000 was calculated to be 1.79 eV based on its optical absorbance onset.Compared to pristine NU-1000, the steady-state photoluminescence (PL) measurement (Figure 1D) showed significantly quenched emission, implying a suppressed electron-hole recombination after TCPP(Co) doping.Meanwhile, gradual blue shift was also observed upon further increasing the TCPP(Co) doping content (Figure S33, more details discussed in Section 16 of SI).In addition, a series of Mott-Schottky experiments at the frequencies of 1000, 1500, and 2000 Hz were carried out (Figure 1E).The positive slopes of the curves agree with the behavior of a typical n-type semiconductor.The flat band potential (E fb ) determined from intersection is −0.8 V versus Ag/AgCl (that is, −0.6 V vs. Normal Hydrogen Electrode [NHE]).Since E fb for a n-type semiconductors is considered to be 0.1-0.2V below the conduction band minimum (CBM), the CBM of TCPP(Co)⊂NU-1000 was estimated to be −0.7 V to −0.8 V versus NHE.Considering the band gap obtained from UV-Vis was calculated to be 1.79 eV, the valence band maximum of TCPP(Co)⊂NU-1000 was estimated to be 1.09V to 0.99 V versus NHE.Compared with the redox potential of CO/CO 2 (−0.53 V vs. NHE, pH 7), these values indicate that the energy band of doped material is suitable for CO 2 photocatalytic reaction.Photocurrent measurements unveil that TCPP(Co)⊂NU-1000 exhibited a higher photocurrent density and lower charge transfer resistance than pristine NU-1000 (Figure 1F, and S18), corresponding to increased chargeseparation efficiency after doping, and is in agreement with the PL results.
The photocatalytic reduction of CO 2 was conducted in aqueous solution under visible-light illumination using triethanolamine (TEOA) as an electron donor.TCPP(Co)⊂NU-1000 exhibited a CO evolution rate of 1314 μmol⋅g −1 ⋅h −1 under light illumination with selectivity as high as 96% (Figure 2A, S19).To the best of our knowledge, TCPP(Co)⊂NU-1000 demonstrates that highest photocatalytic activity and selectivity for CO evolution in aqueous systems without additives such as co-catalysts or photosensitizers (Table S2).The apparent quantum efficiency for CO production under 435-nm irradiation was calculated to be 0.37% (Figure S20 and Equation S2).The isotope labeling experiment using 13 CO 2 (Figure S22) was performed under identical conditions, in which a significant m/z = 29 signal for 13 CO was detected by mass spectra coupled with gas chromatography, confirming the CO produced indeed from CO 2 .In sharp contrast, pristine NU-1000 displays negligible photocatalytic CO production (4.31 μmol⋅g −1 ⋅h −1 ).Control experiments were carried out under Ar, in the absence of TEOA, and in the absence of photocatalyst (Figure 2A), none of which yielded detectable products, implying CO is indeed produced from CO 2 photoreduction.To further understand the role of TCPP(Co), metal free tetra(4-carboxyphenyl)porphyrin (H 2 TCPP) was also tested as dopants in NU-1000, denoted H 2 TCPP⊂NU-1000.Result indicates that it did not produce any CO, CH 4 and H 2 under identical conditions.Additionally, different metalloporphyrins TCPP(Fe) and TCPP(Ni) were also tested as dopants in NU-1000 in the photocatalytic CO 2 reduction and showed poor activities (Figure S24).Thus, it suggests that Co sites played a vital role in the photocatalytic reaction.Additionally, photocatalytic performance strongly depends on the doping content, exhibiting poor activity in higher doping level (Table S1 and Figure S23).Meanwhile, the catalytic activities of TMPP(Co)⊂NU-1000 (introducing tetramethoxyphenylporphyrin cobalt without coordinated groups in NU-1000 by one-pot synthesis), TCPP(Co)-PS⊂NU-1000 (introducing TCPP(Co) in NU-1000 through post-synthesis), and physical mixing of pristine NU-1000 and TCPP(Co) ligand were also evaluated, and these samples showed almost no photocatalytic activity (Figure S24).These results indicate that the dopants need to be incorporated into the crystal lattice, and the chemical bonding between dopant and framework is of vital importance for the FRET process and therefore the CO 2 photoreduction activities.
The photocatalytic reduction of CO 2 in the presence of H 2 O and TEOA was in situ monitored by electron spin resonance (ESR) spectroscopy to elucidate the electron transfer process occurring in NU-1000 and TCPP(Co)⊂NU-1000.Under N 2 atmosphere, the signal at g = 2.005 in NU-1000 was greatly intensified upon visible light illumination, indicating the generation of H 4 TABPy radicals (Figure 2B).However, this signal was much lower in TCPP(Co)⊂NU-1000 and only slightly increase upon light irradiation (Figure 2C, Equation S3). [35,36]This indicated that very few radicals were detected in TCPP(Co)⊂NU-1000 probably due to the efficient energy transfer from H 4 TBAPy to other fragments in the framework.[39][40] As the intensity of Zr 3+ keeps almost unchanged regardless of the environment such as irradiation and atmosphere, we speculated that Zr metal centers are not the active sites for CO 2 reduction.Meanwhile, H 2 TCPP⊂NU-1000 shows barely any catalytic activities, which also indicates Co as the suitable catalytic centers (Figure S25).When CO 2 was introduced into the NU-1000 system, the signal intensity at g = 2.005 increased lower than that in N 2 atmosphere under light irradiation (Equations S3 and S4), suggesting that the electrons derived from H 4 TBAPy were transferred to CO 2 molecules.
In contrast, the intensity of both g = 2.005 and g = 1.968 signals for TCPP(Co)⊂NU-1000 was unchanged under CO 2 no matter with or without light, ruling out the direct charge transfer from H 4 TBAPy to CO 2 or Zr to CO 2 .
In situ DRIFTS were carried out to detect the intermediates produced during photocatalytic CO 2 reduction and understand the possible reaction mechanism.To begin, constant CO 2 flow passing through TEOA aqueous solution was introduced into the catalytic system under darkness to wait for a steady HCO 3 − concentration (Equation 1) and then set as a background for the subsequent measurement.When the light was switched on, the HCO 3 − peak at 1671 cm −1 in the spectra of TCPP(Co)⊂NU-1000 gradually decreased over time (Figure 3B), [41][42][43] while the -OH peak at 3664 cm −1 continued to increase (Figure 3A, Equation 2).These results indicate HCO 3 * was the key intermediate and was transformed into OH − and CO under irradiation. [44,45]To validate this speculation, NaHCO 3 solution was irradiated under the same condition, which gave rise to CO as confirmed by gas chromatograph (GC) analysis (Figure S25).In sharp contrast, pristine NU-1000 shows much lower intensity for the HCO 3 * peak, and no obvious change was observed throughout the entire photocatalytic process (Figure 3C,D).Based on these in situ DRIFTS analyses, we can confirm that doping TCPP(Co) in NU-1000 produced a large amount of intermediates during the reaction, which explains the high photocatalytic activity for CO 2 to CO conversion.
To further understand the relationship between TCPP(Co) molecules and NU-1000, steady state and time-resolved photoluminescence (TRPL) studies were carried out.As shown in Figure 4A, the overlap between the fluorescence emission spectrum of H 4 TBAPy ligand and the absorption spectrum of TCPP(Co) molecules implies the possible energy transfer can occur from H 4 TBAPy to TCPP(Co) in TCPP(Co)⊂NU-1000 (Figure S32).With increasing the TCPP(Co) concentration in H 4 TBAPy solution, the H 4 TBAPy emission peak dramatically decreased accompanied with an obvious red shift approaching to the emission peak of TCPP(Co), also verified by the different doping level in NU-1000 (Figure 4B and S33).Meanwhile, fluorescence quantum-yield (QY) value noticeably decreases from 28.4% to 1.84% after doping (Figure 4C).Since QY results indicate the extent of excited population decays to ground state, therefore the dramatic decreased QY suggest the great excitons migration in doped samples (Figure 4C). [22,46]In addition, TRPL (Figure 4D, Table S3) shows that pristine NU-1000 exhibited an extremely long lifetime of τ = 4.38 ns.In comparison, TCPP(Co)⊂NU-1000 displayed a decay profile with fast decay profile (τ = 1.26 ns), manifesting an obvious lifetime quenching relative to undoped NU-1000. [46]Thus, these results verified the efficient FRET process within TCPP(Co)⊂NU-1000 and a boosted photocatalytic efficiency.Based on these emission lifetimes, the energy transfer rate constant (k ET ) and efficiency (η) were 0.57 ns −1 and 71%, respectively, according to the following equations [47] : Ultrafast transient absorption (TA) spectroscopy characterizations were carried out to track the real-time photoexcited carrier dynamics of pristine NU-1000 and TCPP(Co)⊂NU-1000.TA contour maps of pristine NU-1000 upon photoexcitation (λ ex = 400 nm) feature a negative peak at 445 nm, which can be attributed to ground-state bleaching (GSB) (Figure 4E and S36).The time evolution caused peak broadening, intensity decreased, and red-shifted to 720 nm, which can be assigned to stimulated emission (SE).At the decay time > 900 fs, an excited-state absorption (ESA) peak appeared at ∼770 nm, while the GSB and SE signals gradually decayed.Similarly, TCPP(Co)⊂NU-1000 also displayed GSB peak at 445 nm stemming from H 4 TBAPy* (Figure 4F and S37).However, as the decay of GSB, new peaks around 480, 600, and 770 nm emerged, and the signal intensity gradually enhanced over time (from 150 fs to 400 fs), which is different from pristine NU-1000.According to the UV-Vis absorption and steady-state emission spectra of free TCPP(Co) ligand (Figure S12 and S35), these peaks can be assigned to GSB, S 1 to S 0 SE, and S 1 →S n ESA, respectively, originating from excited-state dynamics of TCPP(Co)* in TCPP(Co)⊂NU-1000.Additionally, if we look into the long-time window up to 10 ps, these TA signals decayed over time without the contributions from H 4 TBAPy excited state (Figure 4G).Hence, it's reasonable to conclude that the signals deriving from TCPP(Co)* were excited by H 4 TBAPy* in TCPP(Co)⊂NU-1000.That is to say, an energy transfer process from H 4 TBAPy to TCPP(Co) in TCPP(Co)⊂NU-1000 occurred after photoexcitation, resulting in TCPP(Co) excited states evolution (more detailed discussion in Section 17 of SI).Furthermore, TCPP(Co)⊂NU-1000 shows about 3-fold increase in peak amplitude (ΔA) at 770 nm compared with NU-1000, indicating a strong S 1 →S n transition dipole moment.This would contribute to photo-excited electrons transfer to higher-lying excited states, which is beneficial to suppress electron-hole recombination, promote excimer formation, and achieve efficient photochemical CO 2 transformation. [48]In addition, the TA kinetics probing at 770 nm (Figure 4H) were fitted by bi-exponential decay functions.The average lifetimes of NU-1000 and TCPP(Co)⊂NU-1000 are 8.92 ps and 30.63 ps (Table S4), respectively.The transient lifetime was increased by more than 3 folds upon introducing TCPP(Co) in NU-1000.The longer decay times in doped materials can be attributed to photoinduced electrons at the conduction band (CB) that can be excited to higher energy levels, leading to enhanced charge carrier separation and photocatalytic performance (Figure 4I and Figure S41). [49]o better understand the energy transfer and electronhole separation behaviors during CO 2 photoreduction in the presence of TCPP(Co)⊂NU-1000, density functional theory (DFT) and time-dependent DFT (TD-DFT) calculations were performed with vienna Ab initio simulation package (VASP) and Gaussian packages.TCPP(Co)⊂NU-1000 model was built by anchoring TCPP(Co) at Zr 6 clusters in pristine NU-1000 and optimized through VASP.The excitation energies and corresponding oscillator strengths of the first 60 and 120 excited states for NU-1000 and TCPP(Co)⊂NU-1000 were investigated, respectively (Table S5 and S6).As shown in Figure 5A, the excitation process of NU-1000 (S 0 →S 25 ) is mainly contributed by H 4 TBAPy units, making it difficult to achieve efficient electron-hole separation.In TCPP(Co)⊂NU-1000, if the FRET process does not occur and only TCPP(Co) units act as photosensitizer, the generated electron-hole pairs would hardly separate according to the S 0 →S 232 transition as shown in Figure 5B S7 in supporting information for details).The larger D index of TCPP(Co)⊂NU-1000 over NU-1000 indicates a farther distance between electron and hole centers, and higher t represents TCPP(Co)⊂NU-1000 shows a higher degree of electron and hole coverage.These results further support that the efficiency of electron-hole separation was significantly improved upon introducing TCPP(Co) into NU-1000.

CONCLUSION
In this work, we have demonstrated that establishing a FRET pathway in NU-1000 through molecular doping can bring significant improvement in catalytic activity for CO 2 photoreduction.Compared with the trace CO production rate over pristine NU-1000, TCPP(Co)⊂NU-1000 shows a CO production rate of 1314 μmol⋅g −1 ⋅h −1 with very high selectivity of 96%, which is among the highest level of reported studies under similar reaction conditions.Experimental and computational studies reveal that the FRET process achieved through TCPP(Co) doping into NU-1000 greatly altered the physicochemical properties of the material and the electron-hole separation efficiency, thus dramatically promoting its photocatalytic performance.Our finding presents a feasible and facile strategy for tuning MOF-based photocatalysts without altering the original structures.

F
I G U R E 4 (A) Emission spectrum of free H 4 TBAPy ligand excited at 375 nm and the UV-Vis absorption of free TCPP(Co) ligand in N,Ndimethylformamide (DMF).(B) Emission spectra of H 4 TBAPy solution with adding different amount of 1 mg/mL TCPP(Co) solution.(C) The quantum yield of pristine NU-1000 and TCPP(Co)⊂NU-1000 excited at 375 nm with H 2 O as background.(D) The time-resolved photoluminescence (PL) kinetics for pristine NU-1000 and TCPP(Co)⊂NU-1000 excited at 375 nm.The time-resolved transient transient absorption (TA) spectra of 1 mg/mL e) pristine NU-1000 and f, g) TCPP(Co)⊂NU-1000 dispersed in (DMF), respectively.h) The TA kinetics curves of pristine NU-1000 and TCPP(Co)⊂NU-1000 dispersed in DMF with the probing wavelength of 780 nm.I) Schematic representation of the photoprocesses in TCPP(Co)⊂NU-1000 upon light irradiation.
. In contrast, the excited processes of S 0 →S 118 and S 0 →S 44 indicate abundant electrons in Co sites could be achieved due to energy transfer between H 4 TBAPy and TCPP(Co) in TCPP(Co)⊂NU-1000 F I G U R E 5 (A) The electron-hole distribution of S 25 in NU-1000.(B-D) The electron-hole distribution of S 232 , S 118 and S 44 in TCPP(Co)⊂NU-1000.Yellow and blue present holes and electrons, respectively.upon light illumination, in agreement with the ESR results (Figure 5C,D).To further clarify and quantify the efficiency of the electron-hole separation, D, t, Sm, and Sr indexes were evaluated (see Table