Graphitic phosphorus coordinated single Fe atoms for hydrogenative transformations

Single-atom metal-nitrogen-carbon (M-N-C) catalysts have sparked intensive interests, however, the development of an atomically dispersed metal-phosphorus-carbon (M-P-C) catalyst has not been achieved, although molecular metal-phosphine complexes have found tremendous applications in homogeneous catalysis. Herein, we successfully construct graphitic phosphorus species coordinated single-atom Fe on P-doped carbon, which display outstanding catalytic performance and reaction generality in the heterogeneous hydrogenation of N-heterocycles, functionalized nitroarenes, and reductive amination reactions, while the corresponding atomically dispersed Fe atoms embedded on N-doped carbon are almost inactive under the same reaction conditions. Furthermore, we find that the catalytic activity of graphitic phosphorus coordinated single-atom Fe sharply decreased when Fe atoms were transformed to Fe clusters/nanoparticles by post-impregnation Fe species. This work can be of fundamental interest for the design of single-atom catalysts by utilizing P atoms as coordination sites as well as of practical use for the application of M-P-C catalysts in heterogeneous catalysis.

energy of -0.407 eV, suggesting that the H 2 can be activated on the Fe-P 4 sites originated from the present M-P-C catalyst. These results clearly demonstrate the unique coordination environment enabled by P compared to that of commonly used N in the hydrogenation reactions. Finally, we also conducted DFT calculations to understand the mechanism of the hydrogenation reaction of quinoline over Fe-P 900 -PCC (Fig. 4).
Accordingly, we have revised our manuscript in Page 4 as the following:   Fig. 14), indicating that the Fe-N 4 sites are intrinsically inert to activate H 2 , which is consistent with the experimental results." The above mentioned Supplementary Fig. 12-14 and Supplementary Table 7  suggesting the single Fe sites in Fe@Fe-N 900 -PCC adopt a planar Fe-N 4 structure (as presented in Supplementary Fig. 13b). b The EXAFS spectrum of Fe-P 900 -PCC shows that the main peak locates at 1.63 Å, ascribing to Fe-P first shell coordination. Furthermore, the Fe-O first shell coordination at 1.45 Å is also included in this broadening peak, which indicates that O need to be included in the curve fitting. On the other hand, a shoulder peak at 2.55 Å for Fe-C second shell coordination is also observed. Therefore, a three-shell structure model, including a Fe-P, a Fe-O and a Fe-C shell, is initially used to fit the EXAFS data of Fe-P 900 -PCC within the R-rang of 1 -3.1 Å and k-rang of 1.42 Å -1 -9.62 Å -1 . The best-fitting analyses manifests that the dominant contribution is originated from Fe-P and Fe-O first shell coordination as presented in Manuscript Fig. 3c and 3d. The coordination numbers for P and O atoms are calculated as 4.0 ± 0.8 and 2.0 ± 0.4, and the corresponding mean bond length of Fe-P and Fe-O are 2.35 ± 0.02 Å and 2.00 ± 0.03 Å, respectively. These results reveal that the single Fe atom in Fe-P 900 -PCC coordinates with four P atoms and a dioxygen molecule (O 2 -Fe-P 4 ). Because the atomic size of P (106 pm) is larger than C (75 pm), Fe center adopts a pyramidal geometry as shown in Manuscript Fig. 3e, this structure is quite different from the planar structure of Fe-N 4 .
Subsequently, we have revised our manuscript in Page 6 as the following: "The coordination information of Fe atom in optimal Fe-P 900 -PCC is further studied using quantitative EXAFS curve fitting (see Supplementary Table 7 for details). The best-fitting analyses are shown in Fig. 3c and 3d, which manifests that the Fe atom are coordinated by four P atoms and a dioxygen molecule (O 2 -Fe-P 4 ), forming a pyramidal geometry as shown in Fig. 3e, this structure is quite different from the planar structure of Fe-N 4 . Besides, the theoretical calculations reveal that the binding energy of O 2 -Fe-P 4 structure is -2.09 eV, suggesting this coordination structure is thermodynamically stable." The revised Fig. 3 has been added into page 6 of the revised manuscript as below: Finally, we have revised our manuscript in Page 7 as below: "DFT calculations. To understand the possible mechanism for the hydrogenation of quinoline, we studied the reaction process with DFT calculations, which initially reveals that the bond length of O  Table 9 and 10). Then, one hydrogen is transferred to N atom (C 9 H 8 N*), leaving another hydrogen bound to Fe atom (int-2 to int-3). Hydrogenation of C 9 H 8 N* to adsorbed C 9 H 9 N* proceeds with an energy barrier of 0.38 eV and is exothermic by -0.348 eV (from int-3 to int-4). The reaction continues to give final product via the addition of second hydrogen molecule, and the reaction of C 9 H 9 N* to C 9 H 10 N* (int-5 to int-6) is found to be the rate determining step with an energy barrier of 0.728 eV." The above mentioned Fig. 4 is added into the revised manuscript as below:

Comment 3. Why the different annealing temperature leads to different Fe content?
Response 3. Thank you very much for this insightful comment. In this work, the porous carbon catalysts were synthesized by a hard-template method. After precursors were pyrolyzed under inert atmosphere, the resultant Fe@carbon@SiO 2 composite need to be washed two times by hydrofluoric acid to remove SiO 2 template. Meanwhile, most of metallic Fe nanoparticles formed in the process of carbonization were also removed by hydrofluoric acid. Therefore, the remained Fe species in the finally obtained catalysts mainly are graphitic phosphorus coordinated single Fe atoms and a certain amount of carbon layers buried Fe nanoparticles. It should be pointed out that the Fe contents of Fe-P x -PCCs (x stands for annealing temperature, 700 to 1100 o C) were characterized by a "combustion method", wherein the catalysts were firstly calcined under air atmosphere, then the obtained residual were dissolved by aqua regia and tested by ICP-MS. In this way, both of graphitic P coordinated single Fe atoms and carbon layers buried Fe nanoparticles are included in the total Fe content. It has been found in the fabrication of Fe-P x -PCCs that increasing the annealing temperature from 700 to 900 o C leads to more insertion of P atoms into the skeleton of graphite to form graphitic P, which could coordinate and stabilize with more Fe atoms, and thus the Fe content of Fe-P 700 -PCC, Fe-P 800 -PCC and Fe-P 900 -PCC increased gradually. After raising the annealing temperature to 1000 and 1100 o C, although the contents of graphitic P and the coordinated single Fe atoms are reduced, these higher annealing temperatures are greatly beneficial to the formation of carbon layers buried Fe nanoparticles 1 , which is the reason why the Fe contents of Fe-P 1000-PCC and Fe-P 1100 -PCC are higher than Fe-P 900 -PCC.   (1) and (2), it can be concluded that the enlarged diffraction angle will leads to the decrease of surface area. Therefore, the heteroatom doping effect is the reason for decrease of the surface area of Fe-N 900 -PCC when compare with Fe-C 900 -PCC 1 .
D is the mean size of the crystalline; k is the Scherrer constant (0.89); λ is the X-ray wavelength; β is the peak width half-height; θ is the Bragg angle; ρ is the crystal density; S is the surface area.
For Fe-P 900 -PCC, the phytic acid was used as phosphorus precursor. In fact, it is also a relatively strong acidic reagent, which will promote the thermal polymerization of carbon precursor (sucrose) and improve the condensation degree of carbon material, just like the role of H 2 SO 4 used in the preparation process of CMK-3 2 . Therefore, it is speculated that the acidity of phytic acid results in the larger pore size and lower specific surface area of Fe-P 900 -PCC compared with corresponding Fe-C 900 -PCC and Fe-N 900 -PCC ( Supplementary Fig. 2b).

Comment 5. A high density of single Fe atoms in Fe-N 900 -PCC, Fe-P 900 -PCC with a very low Fe content may be conflicting. EELS analysis should be provided to confirm the local coordination environment.
Response 5. Thank you very much for the nice suggestion. We are sorry for this conflicting statement caused by the improper use of language and we have corrected this improper expression as the following (Page 3): "As shown in Fig. 1d, Supplementary Fig. 7 and 8, a high density of single Fe atoms is uniformly dispersed on both Fe-P 900 -PCC and Fe-N 900 -PCC." has been changed to "As shown in Fig. 1d, Supplementary Fig. 7, and 8, single Fe atoms are dispersed on these catalysts." EELS is a powerful analytical technique to determine the local coordination state of single metal atom. As your suggestion, we have performed EELS characterization to probe the local coordination information of Fe atoms in Fe-P 900 -PCC. However, because the detection limit of TEM's energy spectrum we utilized is relatively low, the EELS atomic spectrum of Fe element is not pronounced and cannot gain useful information ( Figure R2).   Supplementary Fig. 15). On the other hand, we synthesized a control Fe catalyst, Fe-P 900 -PCCpolymer, through pyrolysis of Fe coordinated P containing organic polymer, and then the same gas phase H-D exchange experiment revealed that the H 2 can be effectively activated over this catalyst ( Supplementary Fig. 15). These two control experiments indicate that the P functional groups cannot catalyze the hydrogenation reaction, and it was speculated that the Fe species interacted with the P species in Fe-P 900 -PCC co-catalyzed the H 2 dissociation and hydrogenation reaction.
The P 2p XPS spectrum of Fe-P 900 -PCC demonstrates that there are three kinds of P functional groups on the surface of Fe-P 900 -PCC: C-O-P, C-PO 2 /C 2 -PO 2 and graphitic P. To determine which P group coordinated Fe is the active sites, we then fabricated a series of Fe-P x -PCC (x stands for carbonization temperature, 700-1100 o C). It was revealed that the content of graphitic P is positively associated with the catalytic activities, which might mean graphitic P coordinated Fe is the active sites for hydrogenation reaction, rather than graphitic P itself, because above H-D exchange experiment has approved the P-doped carbon catalyst cannot disassociate the H 2 .
This speculation is further studied by different characterizations. It has been revealed in Fe-P 700 -PCC that P atoms could not be inserted into the graphitic carbon framework at 700 o C, all the P species are C-O-P and C-PO 2 /C 2 -PO 2 in the carbon surface, these P species cannot form stable coordination or interaction structure with Fe species, then the vast majority of Fe were washed off by hydrofluoric acid at template removing step, as expected, Fe-P 700 -PCC was also inactive for hydrogenation reaction. When raising the carbonization temperature to above 800 o C, XPS and solidstate 31 P NMR spectra indicate that part of P atoms has been inserted into the carbon framework to form graphitic phosphorus (0.21 atomic %). Furthermore, the EXAFS results also confirm the formation of Fe-P coordination. Gratifyingly, Fe-P 800 -PCC gave a 69% conversion of quinoline.
Elevating carbonization temperature to 900 o C produced the highest content of graphitic phosphorus

Comment 7. Moreover, the fitted P XPS spectra showed that no Fe-P state while XAFS results
presented Fe-P state, it is contradictory.

Response 7.
Thank you for this comment. In the P 2p XPS spectrum, the binding energy of Fe-P bond is in the range of 129-130 eV 1-2 . In this work, the relatively low content of Fe in the catalyst results in less Fe-P coordination on the surface of catalyst, and the content of Fe-P is lower than the detection limit of XPS, so it is not observed in the XPS analysis (as shown in Figure R3). On the other hand, the Fe k-edge EXAFS is for characterizing the coordination atoms of Fe and the detection limit of XAFS is much lower than XPS, so that Fe-P coordination could be detected by XAFS.

Figure R3
The P 2p XPS spectrum of Fe-P 900 -PCC.  Fig. 18a and 18b), and the content of graphitic P in Fe-P 900 -PCC remains unchanged before and after hydrogenations (0.41 vs 0.41 atomic %, Supplementary  Fig. 18c and 18d). All of these results demonstrate that the single Fe sites over Fe-P 900 -PCC are stable under reaction conditions. Accordingly, we have added below sentence into page 10 of the revised manuscript:

Comment 8. And the content of P in various samples should be measured by ICP, because of the incorrect content measured by XPS. It is complex that the content of P presented atomic % and
"P 2p and Fe 2p XPS spectra reveal that the chemical state of P grap and Fe species keep constant ( Supplementary Fig. 18 a and b), the content of P grap remains unchanged at 0.41 atomic % before and after hydrogenations (Supplementary Table 6). The AC-STEM image and Fe k-edge EXAFS spectrum of spent Fe-P 900 -PCC indicate that the atomically dispersed Fe species are well preserved after eight repetitive runs (Supplementary Fig. 18c and d).  Tables   To further understand the hydrogenation reaction process at the atomic level, DFT calculations were performed to study the reaction mechanism (hydrogenation of quinoline was choose as model reaction Accordingly, we have revised our manuscript in Page 6 as the following:

Comment 10. A comprehensive comparison with relative reports should be in
"The coordination information of Fe atom in optimal Fe-P 900 -PCC is further studied using quantitative EXAFS curve fitting (see Supplementary Table 7 for details). The best-fitting analyses is shown in Fig. 3c and 3d, which manifest that the Fe atom is coordinated by four P atoms and a dioxygen molecule (O 2 -Fe-P 4 ), forming a pyramidal geometry as shown in Fig. 3e, this structure is quite different from the planar structure of Fe-N 4 . Besides, the theoretical calculations reveal that the binding energy of O 2 -Fe-P 4 structure is -2.09 eV, suggesting this coordination structure is thermodynamically stable." The revised Fig. 3 has been added into page 6 of the revised manuscript as below:  Fig. 13 and Manuscript   Fig. 3).CN, coordination number; R, distance between absorber and backscatter atoms; σ 2 , the Debye-Waller factor; ΔE 0 , inner potential correction; R-factor, indicate the goodness of the fit. a For the EXAFS spectrum of Fe@Fe-N 900 -PCC ( Supplementary Fig. 12), only a strong Fe-N peak at 1.51 Å is observed. So, the fitting was performed by including a single Fe-N shell within the R-rang of 1 -3.1 Å and k-rang of 1.42 Å -1 -9.62 Å -1 . The fitting results reveal that the coordination number of Fe center with surrounding N atoms is 4.2 ± 0.4 and the average Fe-N bond length is 1.95 Å ± 0.01, suggesting the single Fe sites in Fe@Fe-N 900 -PCC adopt a planar Fe-N 4 structure (as presented in Supplementary Fig. 13b).
b The EXAFS spectrum of Fe-P 900 -PCC shows that the main peak locates at 1.63 Å, ascribing to Fe-P first shell coordination. Furthermore, the Fe-O first shell coordination at 1.45 Å is also included in this broadening peak, which indicates that O need to be included in the curve fitting. On the other hand, a shoulder peak at 2.55 Å for Fe-C second shell coordination is also observed. Therefore, a three-shell structure model, including a Fe-P, a Fe-O and a Fe-C shell, is initially used to fit the EXAFS data of  Table 9 and 10). Then, one hydrogen is transferred to N atom (C 9 H 8 N*), leaving another hydrogen bound to Fe atom (int-2 to int-3). Hydrogenation of C 9 H 8 N* to adsorbed C 9 H 9 N* proceeds with an energy barrier of 0.38 eV and is exothermic by -0.348 eV (from int-3 to int-4). The reaction continues to give final product via the addition of second hydrogen molecule, and the reaction of C 9 H 9 N* to C 9 H 10 N* (int-5 to int-6) is found to be the rate determining step with an energy barrier of 0.728 eV." The above mentioned Fig. 4 is added into the revised manuscript as below: Fig. 4 Energy profile of hydrogenation of quinoline over Fe-P 900 -PCC.
Supplementary Table 9 and 10 have been added in Supplementary Information: Table 9.

Supplementary
Step by step barrier (E a , eV) and reaction energy (E r , eV) for hydrogenation of quinoline (C 9 H 7 N) over Fe-P 900 -PCC. Response. We appreciate the reviewer very much for the positive comment.

Fe-P-C catalyst. What are the coordination numbers of Fe-P? How Fe is coordinated with graphic P? What is the structure?
Response 1. We appreciate this insightful comment very much. Following your and other reviewer's advice, the atomic structure of active sites in Fe-P 900 -PCC has been investigated through quantitative EXAFS curve fitting and theoretical calculations.  Table 7). These results revealed that the single Fe sites in Fe-P 900 -PCC are coordinated with four P atoms and a dioxygen molecule (O 2 -Fe-P 4 ). Because the atomic size of P (106 pm) is larger than C (75 pm), Fe center adopts a pyramidal geometry as shown in Fig. 3e. Besides, the theoretical calculations reveal that the binding energy of O 2 -Fe-P 4 structure is -2.09 eV, suggesting this coordination structure is thermodynamically stable.
Subsequently, we have revised our manuscript in Page 6 as the following: "The coordination information of Fe atom in optimal Fe-P 900 -PCC is further studied using quantitative EXAFS curve fitting (see Supplementary Table 7 for details). The best-fitting analyses is shown in Fig. 3c and 3d, which manifest that the Fe atom is coordinated by four P atoms and a dioxygen molecule (O 2 -Fe-P 4 ), forming a pyramidal geometry as shown in Fig. 3e, this structure is quite different from the planar structure of Fe-N 4 . Besides, the theoretical calculations reveal that the binding energy of O 2 -Fe-P 4 structure is -2.09 eV, suggesting this coordination structure is thermodynamically stable." The revised Fig. 3 has been added into page 6 of the revised manuscript as below:   Fig. 13 and Manuscript   Fig. 3).CN, coordination number; R, distance between absorber and backscatter atoms; σ 2 , the Debye-Waller factor; ΔE 0 , inner potential correction; R-factor, indicate the goodness of the fit.
a For the EXAFS spectrum of Fe@Fe-N 900 -PCC ( Supplementary Fig. 12), only a strong Fe-N peak at 1.51 Å is observed. So, the fitting was performed by including a single Fe-N shell within the R-rang of 1 -3.1 Å and k-rang of 1.42 Å -1 -9.62 Å -1 . The fitting results reveal that the coordination number of Fe center with surrounding N atoms is 4.2 ± 0.4 and the average Fe-N bond length is 1.95 Å ± 0.01, suggesting the single Fe sites in Fe@Fe-N 900 -PCC adopt a planar Fe-N 4 structure (as presented in Supplementary Fig. 13b).
b The EXAFS spectrum of Fe-P 900 -PCC shows that the main peak locates at 1.63 Å, ascribing to Fe-P first shell coordination. Furthermore, the Fe-O first shell coordination at 1.45 Å is also included in this broadening peak, which indicates that O need to be included in the curve fitting. On the other hand, a shoulder peak at 2.55 Å for Fe-C second shell coordination is also observed. Therefore, a three-shell structure model, including a Fe-P, a Fe-O and a Fe-C shell, is initially used to fit the EXAFS data of We have revised our manuscript in Page 7 as below: "DFT calculations. To understand the possible mechanism for the hydrogenation of quinoline, we studied the reaction process with DFT calculations, which initially reveals that the bond length of O  Table 9 and 10). Then, one hydrogen is transferred to N atom (C 9 H 8 N*), leaving another hydrogen bound to Fe atom (int-2 to int-3). Hydrogenation of C 9 H 8 N* to adsorbed C 9 H 9 N* proceeds with an energy barrier of 0.38 eV and is exothermic by -0.348 eV (from int-3 to int-4). The reaction continues to give final product via the addition of second hydrogen molecule, and the reaction of C 9 H 9 N* to C 9 H 10 N* (int-5 to int-6) is found to be the rate determining step with an energy barrier of 0.728 eV." The above mentioned Fig. 4 is added into the revised manuscript as below: Fig. 4 Energy profile of hydrogenation of quinoline over Fe-P 900 -PCC.
Supplementary Table 9 and 10 have been added in Supplementary Information: Table 9.

Supplementary
Step by step barrier (E a , eV) and reaction energy (E r , eV) for hydrogenation of quinoline (C 9 H 7 N) over Fe-P 900 -PCC.

Comment 3. On page 6, the authors claimed that "the formation of P grap is the prerequisite for the existence of single Fe atoms on the surface of P-doped porous carbon". What is the evidence for this? Then what is the form of the Fe in Fe-P 700 -PCC?
Response 3. Thank you very much for this valuable comment. Your concern on the relationships between graphitic phosphorus and single Fe atoms is important, which helps us to give a clearer explanation for the formation mechanism of active sites.
At 700 o C, P atoms cannot insert into graphitic carbon framework, all of the P species are PO x (determined by XPS and solid-state 31 P NMR), these P functional groups are unstable and cannot form stable structure with Fe atoms, the vast majority of Fe came from raw materials were removed by hydrofluoric acid at template removing step. When increasing the carbonization temperature to above For Fe-P 700 -PCC, we have intended to investigate the chemical state of Fe species in this sample through X-ray absorption spectroscopy (XAS). Unfortunately, since the Fe content of Fe-P 700 -PCC is very low (0.0072 wt%), the signal of Fe in X-ray absorption near edge structure (XANES) is very weak and cannot get useful information from this characterization ( Figure R1). The TEM characterizations indicate that there are a number of Fe nanoparticles on the surface of Fe 1.0 /Fe-C 900 -PCC and Fe 1.0 /Fe-N 900 -PCC ( Figure R3). Then, their catalytic performances in the hydrogenations of quinoline and nitrobenzene have been assessed. As listed in Table R1 and Table R2, both of them have very low catalytic activities in these two hydrogenation reactions, which are similar to that of Fe 0.95 /Fe-P 900 -PCC.   The conversion and yield were determined by GC using n-hexadecane as an internal standard.