Exploring Spin Distribution and Electronic Properties in FeN4-Graphene Catalysts with Edge Terminations
Round 1
Reviewer 1 Report
Comments and Suggestions for AuthorsThe work is devoted to the study of Electronic Properties in FeN4-2 Graphene. The research is interesting because it aims to create catalysts without platinum. The authors study the influence of FeN4 defects on the properties of the system. The authors illustrate the results of their research in detail. Additional material is also provided in the SI file. The list of references includes 64 publications from 2006 to 2023. The authors study a system that contains iron. Such systems are quite difficult to model. Therefore, there are several questions about the correctness of the calculation methods used.
1. The system contains iron. Is DFT+U used for calculations of this kind? Why was this approximation not used?
2. Why is 15x1x1 and 1x15x1 grid used? Do I understand correctly that 15 points are taken along one direction? Why isn't 15x15x1 used? Is such a grid sufficient? Conductors often require a very large number of k-points.
Author Response
We thank the reviewer for his/her evaluation of our manuscript.
Reviewer comment
- The system contains iron. Is DFT+U used for calculations of this kind? Why was this approximation not used?
Author’s reply:
In our models, the Fe-N-C system contains a single Fe atom, leading to limited d electronic interactions. This decreases the need for the Hubbard U correction in our model. Instead, we used low smearing parameters to enhance the visibility of bands near the Fermi level without the complexity of DFT+U.
Reviewer comment
- Why is 15x1x1 and 1x15x1 grid used? Do I understand correctly that 15 points are taken along one direction? Why isn't 15x15x1 used? Is such a grid sufficient? Conductors often require a very large number of k-points.
Author’s reply:
In the graphene nano-ribbons, the periodicity of edge configurations is established along one direction. Zigzag edge configurations are modelled with periodicity in the x-direction, while armchair configurations are modelled with periodicity in the y-direction. For this reason, k-points along the periodicity directions play a role in the computations. The optimized grid number of k-points we selected exceed the k-point grid commonly used in similar studies that employ VASP calculations. We have conducted extensive optimizations of k-points and cut-off energy to choose the appropriate grid parameters. Increasing the k-points up to 20x1x1 and 1x20x1 did not significantly change the energy values, confirming the adequacy of the chosen grid.
Action taken:
In p. 11 we added “Increasing the k-points up to 20 along the periodicity direction did not significantly change the energy values, confirming the adequacy of the chosen grid.”
Reviewer 2 Report
Comments and Suggestions for AuthorsIn this manuscript, the authors studied the spin distribution in FeN4-doped graphene nanoribbons with zigzag and armchair terminations. This study employs a periodic DFT-based method to systematically evaluate the variation of spin-moment distribution and electronic properties based on the position, orientations of FeN4 defects, and the edge terminations of graphene nanoribbons, which reveals that electronic structures in both zigzag and armchair geometries are highly sensitive to the location of FeN4 defects. Additionally, the introduction of FeN4 defects at edge positions eliminates the known dependence of magnetic and electronic properties on undoped graphene nanoribbons' edge geometries.
This work can be interesting to both computational and experiment chemistry community. As such, the proposed article deserves to be published and the Molecules is certainly well targeted.
Before the publication, I would like to ask the authors to consider the minor comments below.
1. Page 4, Figure 2
The binding energy of D is significantly higher than D1, which is different from other pairs (A-A1, B-B1 and C-C1), where the binding energy of the bond in the horizontal direction is higher, could the authors provide more discussion on this issue?
2. Page 7, Figure 7
Band gaps shown in this picture are obtained from PBE, which is known to largely underestimate band gaps in periodic systems. I suggest mentioning this issue in the manuscript: band gaps from PBE can only be used for qualitative comparison, for quantitative comparison, range-separated functionals or GW approximation is needed.
3. Page 9, Figure 5
As shown in Figure 5, the energy of the dz2 and dx2-yz bands do not change with respect to the k-point. Especially for the dz2 band, it almost completely overlaps with the Fermi level. Can the authors provide more explanations for this?
4. Page 10, Computational details
Was there finite-temperature smearing used in the DFT calculations of the AFM phase?
5. Page 10, line 360
“The exchange–correlation energy was calculated within the Perdew Burke and Ernzerhof formulation of the generalized-gradient approximation (GGA-PBE)”
For the systems studied in this work, the long-range dispersion interaction might be important. Can the authors briefly discuss the impact of not adding the dispersion correction in the PBE functional?
Comments on the Quality of English LanguageNo major language or grammar problem found.
Author Response
We thank the reviewer for his/her evaluation of our manuscript.
Reviewer comment
- Page 4, Figure 2
The binding energy of D is significantly higher than D1, which is different from other pairs (A-A1, B-B1 and C-C1), where the binding energy of the bond in the horizontal direction is higher, could the authors provide more discussion on this issue?
Author’s reply:
In both edge models, the binding energy of the Fe atom on edge pyridinic-N defect configurations is higher than that of other configurations. he Fe binding energy of configuration D1 is higher than that of D, consistent with the equivalent corresponding configuration in zigzag (C1-C). This can be explained with the changes of the carbon matrix environment around FeN4 when moving it from in-plane to the edges.
Action taken:
We have added the following text to discuss the differences between D and D1 locations in p. 5 :
“This can be associated with the different FeN4 environments in D and D1 locations, which apparently contribute stronger to the FeN4 stabilization than the specific orientation of FeN4. In the D-placed defect with “tilted” orientation one nitrogen becomes pyrrolic-like (at the edge) and the other three nitrogens remain pyridinic, whereas in the D1-placed defect two nitrogens are pyrrolic-like. The D1-defect has the same carbon environment as the most stable C1-defect in ZGNR with two pyridinic and two pyrrolic-like nitrogens. Thus, increasing the number of the pyrrolic-like nitrogen atoms at the edges increases the thermodynamic stability of FeN4 -GNRs. This aligns with Li et al.'s findings [28], where nitrogen defects near the edges of ZGNRs are energetically favorable, and pyrrole-like defects possess even lower formation energies than pyridinic-like defects. Our findings on Fe-centered defect stabilization in graphene nanoribbons are consistent with Holby et al.'s observation [29] of preferential edge stabilization of Fe-pyridinic vacancy complexes (FeN3) in GNRs.”
Reviewer comment
- Page 7, Figure 7
Band gaps shown in this picture are obtained from PBE, which is known to largely underestimate band gaps in periodic systems. I suggest mentioning this issue in the manuscript: band gaps from PBE can only be used for qualitative comparison, for quantitative comparison, range-separated functionals or GW approximation is needed.
Author’s reply:
We agree that GGA often underestimates band gaps in systems with strongly correlated electrons as are the systems with 3d elements and that using GW is the best choice. In our model, there is only one Fe atom with 3d electrons and this limits the correlation effects. In addition, we carried out a comparative study on structures with same chemical compositions and therefore, we expect that the errors due to GGA-PBE functional are of the same order across all the structures.
Action taken : We added the following text on page 7, below Fig. 4 :
« The band-gap values presented in Figure 4b,d are calculated using the GGA-PBE functional, known to underestimate band gaps, especially for systems with strong electronic correlations. However, while PBE may not yield precise absolute band gap values, it reliably captures trends in the band structures variations with FeN4 location and orientation.”
Reviewer comment
- Page 9, Figure 5
As shown in Figure 5, the energy of the dz2 and dx2-yz bands do not change with respect to the k-point. Especially for the dz2 band, it almost completely overlaps with the Fermi level. Can the authors provide more explanations for this?
Author’s reply:
In our analysis with the partial density of states added to the band diagram (fat-band), we observed a localized flat band characteristic of d(x2-y2) and dz2. The contributions of dyz and dxz in the band are intertwined with the pz orbital contribution from carbon. This results in hybridization between the C2pz and Fe 3dxz/3dyz orbitals.
Action taken:
- We replaced Figure 4 with a new Figure to indicate the d-states of interest
- The Figure caption of Figure 4 was changed by adding the explanation of the arrows: “In (b), the blue arrows point to the dz2 states and the green arrows point to the dx2-y2 (“direct” orientation) or dxy (“tilted” orientation) states”
- we provided the following explanations in the revised manuscript, page 8:
“Localized flat band characteristics due to dx2-y2, dxy and dz2 states are observed in the band structures in all FeN4-ZGNRs as shown by the blue and green arrows pointing to these d-bands in Figure 4b. The unoccupied dz2 localized band (pointed by a blue arrow in Figure 4b) shifts downward the Fermi level in moving the FeN4 defect from A1 to B1, to C1. The dz2 unoccupied band localizes at energy levels of 0.34, 0.21, and 0.01 eV in A1, B1 and C1 configurations, respectively. The contributions of dyz and dxz in the bands are intertwined with the pz orbital contribution from carbon. This results in hybridization between the Cpz and Fe dxz/dyz orbitals, as demonstrated for C1 FeN4-ZGNR structure in Figure 5.”
Reviewer comment
- Page 10, Computational details
Was there finite-temperature smearing used in the DFT calculations of the AFM phase?
Author’s reply:
We consistently employed the Fermi-Dirac smearing method across all cases. Specifically, we set the smearing parameter (sigma) to 0.03, which was found to be appropriate for our calculations.
Action taken:
We added on page 11 the following:
“We consistently employed the Fermi-Dirac smearing method across all cases, setting the smearing parameter to 0.03.”
Reviewer comment
- Page 10, line 360
“The exchange–correlation energy was calculated within the Perdew Burke and Ernzerhof formulation of the generalized-gradient approximation (GGA-PBE)”. For the systems studied in this work, the long-range dispersion interaction might be important. Can the authors briefly discuss the impact of not adding the dispersion correction in the PBE functional?
Author reply’s
While it's true that including dispersion corrections in the PBE functional could quantitatively affect certain properties, such as the binding energy of the Fe atom, we anticipate that this would apply uniformly across the different configurations. Therefore, it is unlikely to alter the relative trends and comparative analyses that form the core of our study. The primary focus of our research is on understanding these trends, particularly in terms of the position of the Fe-N4 centre within the graphene matrix.
Round 2
Reviewer 1 Report
Comments and Suggestions for AuthorsAccept
Reviewer 2 Report
Comments and Suggestions for AuthorsThe authors have correctly addressed all my technical remarks. Further review is not needed.
Comments on the Quality of English LanguageNo major language issue found.
