Enhanced Hydrogen Evolution Performance of Carbon Nitride Using Transition Metal and Boron Co‐Dopants

Density functional theory calculations are used to study the effect of several metal dopants (M = Ag, Cd, Co, Cu, Fe, Ni, Pt, Sc, Ti, and Zn) and metal–boron co‐dopants on the structure and catalytic property of g‐C3N4 2D monolayer. Using transition metals and boron (TM–B) as co‐dopants not only keeps the 2D structure stability of g‐C3N4 monolayer, but also alters the catalytic performance of the structures. The co‐doping of B in TM (TM = Pt, Zn, Cd, Ti, and Sc)‐doped g‐C3N4 leads to a significant increase in the hydrogen adsorption energy because hydrogen binding site changes from N to C. For TM–B (TM = Fe, Co, and Ni) co‐doped g‐C3N4, the hydrogen adsorption energy has no obvious change since the hydrogen binding site remains on C atom near the doped TM. However, the co‐doping of B in TM‐ (TM = Cu and Ag) doped g‐C3N4 leads to a significant reduction of hydrogen adsorption energy, making them good candidates for hydrogen evolution reaction. This study provides theoretical guidance for the experimental synthesis of TM–B co‐doped g‐C3N4 and paves a way for the design of a widely applicable non‐noble catalyst.

DOI: 10.1002/sstr.202200264 Density functional theory calculations are used to study the effect of several metal dopants (M ¼ Ag, Cd, Co, Cu, Fe, Ni, Pt, Sc, Ti, and Zn) and metal-boron codopants on the structure and catalytic property of g-C 3 N 4 2D monolayer. Using transition metals and boron (TM-B) as co-dopants not only keeps the 2D structure stability of g-C 3 N 4 monolayer, but also alters the catalytic performance of the structures. The co-doping of B in TM (TM ¼ Pt, Zn, Cd, Ti, and Sc)-doped g-C 3 N 4 leads to a significant increase in the hydrogen adsorption energy because hydrogen binding site changes from N to C. For TM-B (TM ¼ Fe, Co, and Ni) co-doped g-C 3 N 4 , the hydrogen adsorption energy has no obvious change since the hydrogen binding site remains on C atom near the doped TM. However, the co-doping of B in TM-(TM ¼ Cu and Ag) doped g-C 3 N 4 leads to a significant reduction of hydrogen adsorption energy, making them good candidates for hydrogen evolution reaction. This study provides theoretical guidance for the experimental synthesis of TM-B co-doped g-C 3 N 4 and paves a way for the design of a widely applicable non-noble catalyst.
these structures is another approach to tune the electronic properties of g-C 3 N 4 and provide more flexible tunability in these graphene-like structures. [3b,16] While many researchers have attempted to introduce efficient structures to improve the selectivity of different reactions, the lack of scientific explanation has led to poor reproducibility in the field, which has slowed the progress in the field.
Computational tools such as density functional theory (DFT) calculations can help to understand the role of different dopants and co-dopants on catalyst performance in most crystalline and periodic structures. [3b,5b,16c] Without conducting timeconsuming and expensive experiments, DFT results can help to estimate the performance of doped/co-doped structures. [17] So far, many research groups have carried out theoretical studies to study the catalytic properties of the pure and doped g-C 3 N 4 structures. [3a,11c] Jiang et. al. studied the effect of Ag-doped g-C 3 N 4 structure on HER. They compared the effect of doped Ag particle size on HER and found that single Ag atom doping can improve the HER efficiency of g-C 3 N 4 . [11b] In another study, experimental research was carried out to investigate HER of single Pt-doped g-C 3 N 4 . They detected C-TM binding and their DFT calculations showed that the TM-C binding forms in a defective g-C 3 N 4 structure. [18] However, most of the computational research were carried out to explain the experimental results and there are limited systematic computational studies on TM doping and their effect on the HER activity of g-C 3 N 4 . Therefore, it is important to conduct systematic first-principles theoretical studies to design TM-doped or co-doped g-C 3 N 4 -based structures for HER applications.
This study is to use first-principles theoretical calculations to understand the catalytic performance of the TM-doped and TM-B co-doped g-C 3 N 4 structures. By comparing the stability of all the possible doping/co-doping structures and comparing their energies, we identified the most stable structures for TM-doped and TM-B co-doped g-C 3 N 4 . It is found that the TM-doped g-C 3 N 4 becomes greatly deformed, while the TM-B co-doped g-C 3 N 4 can keep the flat 2D structure. We also studied hydrogen adsorption on the single doped and co-doped g-C 3 N 4 and found that the most stable hydrogen adsorption site on the doped/codoped g-C 3 N 4 structures varies with the dopants. Cu-B co-doped structure has comparable hydrogen adsorption energy to that of Pt-doped g-C 3 N 4 , indicating it is a good candidate of non-noble single atom catalyst. The results can be used to design stable and inexpensive co-doped g-C 3 N 4 for hydrogen production in future.

Computational Parameters
DFT as implemented in the Vienna Ab Initio Package (VASP) is applied for all the calculations in this study. [19] The generalized gradient approximation (GGA) with the use of Perdew-Burke-Ernzehof (PBE) functional is applied to adopt the exchangecorrelation effect in all the calculations. [20] To correct the van der Waals interactions, a Becke-Johnson (D3-BJ) dispersion correction is applied. [21] To describe the core and valence electrons, the projector augmented-wave (PAW) method with an energy cut-off of 520 eV and a force tolerance of 10 À4 eV Å À1 is applied.
In this study to calculate the HER efficiency, a supercell of g-C 3 N 4 monolayer containing 112 atoms is modeled. The dimensions of the relaxed g-C 3 N 4 structure for this study are 14.26 Å Â 24.69 Å at X and Y directions, respectively. To minimize the ionic interactions of neighboring layers, a vacuum of 22 Å in Z direction is considered. The following equation is applied to calculate the formation energy of the dopants in the structure of the single-doped g-C 3 N 4 structures.
where E f is the formation energy, E srf þd is the total energy of the doped g-C 3 N 4 , E srf is the total energy of the undoped g-C 3 N 4 , and μ d is the chemical potential of crystalline TMs. The following equation is used to calculate the formation energies of the B-TM co-doped g-C 3 N 4 structures where E srf þdþB is the total energy of the co-doped structure, and μ B and μ C are the chemical potential of boron and carbon atoms, respectively. For hydrogen adsorption on g-C 3 N 4 monolayer, the chemisorption energy of hydrogen is calculated as follows where E srf þH is the total energy of the doped and co-doped g-C 3 N 4 when hydrogen is adsorbed on the surface, E srf is the total energy of doped and co-doped g-C 3 N 4 in the absence of hydrogen, and 1 2 μ H 2 is the chemical potential of one hydrogen atom.

Calculation of the HER Efficiency
The HER is usually defined as the reaction of 2 protonated hydrogen ions and 2 electrons that can produce the hydrogen gas H 2 .
2H þ þ 2e À ! H 2 (4) There are two possible pathways to generate H 2 . the main step is the adsorption of hydrogen on the electrode surface to form H Ã .
The first mechanism happens following this step. If the second hydrogen is adsorbed close to the first hydrogen on the surface, the hydrogens can interact and consequently can form the H 2 molecule.
And the second mechanism happens when the second hydrogen comes from the solution and interacts with the adsorbed H Ã to form H 2 .
After the adsorption of the first hydrogen (Equation (5)), any of the mechanisms (Equation (6) or (7)) can be activated depending on the condition. As the reaction at room temperature is important for the mechanism, the Gibbs free energy should be considered to www.advancedsciencenews.com www.small-structures.com calculate HER. The HER is a reversible spontaneous process meaning that the free energy of H 2 gas is ΔG H Ã ¼ 0. To calculate the energy barrier for HER reaction, the reference energy is set at zero (U 0 ¼ 0). [5b] In this study, the computational hydrogen electrode method is applied to calculate the Gibbs free energy To calculate the Gibbs free energy, the zero-point energy correction ðΔE ZPE Þ and energy contributed by entropy change at temperature of T (TΔS H ) are considered. Here, ΔE ZPE is the energy difference between the absorbed and gas phase hydrogen at zero Kelvin temperature. TΔS H is the entropy term for the gas phase hydrogen at the standard condition. The zero point energy correction has been calculated before and its value is about 0.24 eV.
[5b] It deserves to mention that surface charge is also an important factor that can have an impact on electrochemical reactions. Previously, Kim et al. studied the charge effect of 2D and 3D materials and showed that the effect is even stronger for 2D materials. [22] However, considering the surface charge effect in DFT calculations is very complicated and challenging. In this study, the surface charge effect is neglected, and the standard computational hydrogen electrode model (SCHE) is considered to study the HER. SCHE simplifies the electrochemical interface and ignores the effects of surface charge and interface. [22,23] In this article, VESTA software is employed to visualise the structures. [24] Bader charge analysis is conducted using the code developed by Henkelman group. [25] 3. Results and Discussion

TM Doping of g-C 3 N 4 Monolayer
In this article, the HER on the surface of the M-doped (M ¼ Ag, Cd, Co, Cu, Fe, Pt, Ni, Sc, Ti, and Zn) g-C 3 N 4 monolayer and B-TM (boron-transition metal) co-doped g-C 3 N 4 monolayer is investigated. The main aim is to better understand the role of the TMs as dopants on hydrogen adsorption on g-C 3 N 4 monolayer and to compare the results with the B-TM metal co-doped structures. Most of the TMs considered in this study are nonnoble TM, with the exception of Pt and Ag for comparison.
A relatively large supercell with 112 atoms is selected to eliminate the interactions between neighboring dopants. The g-C 3 N 4 structure is optimized for the creation of doping structures. To find the most stable site for the single metal dopants in the structure of the g-C 3 N 4 monolayer, 13 initial doping sites are considered for each of the dopants. Figure 1a shows the top and side view of the most stable structure of the single metal-doped structures. The most stable site for the TM dopant is at the corner of the triangular g-C 3 N 4 unit, and the C─N bond is broken to accommodate the TM dopant. Meanwhile, the TM atom also binds with two pyridinic N atoms nearby. Such a doping lead to significant structure deformation of g-C 3 N 4 and we can clearly see the crumble of the structure from the side view of Figure 1a. The structure parameters and formation energy of g-C 3 N 4 with various TM dopants are shown in Table 1. The E f values change with the type of TM dopants, varying from À2.061 eV (Ag) to À5.399 eV (Sc).
To further tune the electronic and catalyst properties of g-C 3 N 4 monolayer, introducing boron as a co-dopant in the structure of g-C 3 N 4 monolayer is also considered in our study. The reason for choosing boron is that boron has shown ability to change the  Table 1. The bond lengths of the TM with the surrounded carbon and nitrogen atoms and the formation energy of the TMs in the g-C 3 N 4 monolayer. M refers to the TMs and N1, N2, N3, and C are surrounding nitrogens and carbon atom, respectively (see Figure 1). electronic properties of g-C 3 N 4 and accordingly affect the HER. [26] First, the most stable boron-doped g-C 3 N 4 structure is identified by considering possible interstitial and substitutional doping sites (5 substitutional sites and 13 interstitial sites). The most stable boron-doped structure is used to investigate the stability of the co-doped g-C 3 N 4 structures. In the structure of the boron-doped g-C 3 N 4 , eight possible co-doping sites for the TMs are considered to find the most stable co-doped g-C 3 N 4 structures. All the structures are optimized, and the most stable structures are selected to study HER. Figure 1b shows the most stable co-doped g-C 3 N 4 samples. For all the B-TM co-doped structures, it is found that the doping site of the TMs in the co-doped structures is different from that of the single-doped structures.
In the structure of the single-metal-doped samples, the dopant breaks the carbon─nitrogen bond and form bonding with both carbon and nitrogen atoms (Figure 1a). The side view (Figure 1a) of the doped structure shows a considerable disorder in the structure of the 2D g-C 3 N 4 monolayer. The level of strain in the single metal-doped structures suggests that the doped structures are fragile. Therefore, to keep the structures stable, the doping www.advancedsciencenews.com www.small-structures.com concentration should be controlled. In the structure of B-TM co-doped structures, however, metals prefer to sit in the holes (Figure 1b). Such a doping site is usually considered for single metal-doped structures by many computational researchers. [9c,27] The co-doped g-C 3 N 4 structure is more stable than the singlemetal-doped g-C 3 N 4 structure, and thus to be more feasible for applications such as HER.
The doped and co-doped structures are expected to have different electronic and catalytic properties. To understand the difference between the doped and co-doped g-C 3 N 4 structures, the electronic structure of the doped/co-doped samples is compared. By comparing the density of states (DOS) of the pure and single-metal-doped structures, (Figure 2a,b) we can see that dopants can change the electronic structure of the g-C 3 N 4 monolayer. While most of the doped structures have n-type electronic structure, dopants with unoccupied orbitals (B, Sc, and Ti) can change the g-C 3 N 4 to a p-type semiconductor. For electrocatalyst and photocatalyst purposes, the changes in DOS around the conduction band minimum (CBM) and valance band maximum (VBM) or the fermi energy level are important to the catalytic properties of materials. As the catalytic interaction mainly involves the gain or loss of electrons, the relative position of the Fermi level and CBM determine how easy it is to gain or lose electrons. [28] The results (Figure 2c,d) show that boron can cause a significant change in the electronic structure of g-C 3 N 4 , which agrees with previous reports. [10,14b] After the introduction of boron atom to the single-metal-doped g-C 3 N 4 structures, most of the structures have p-type electronic structure, which is consistent with the shift of the Fermi level.

Hydrogen Adsorption
To study HER on the TM-doped and TM-B co-doped g-C 3 N 4 samples, the most stable structures identified by the aforementioned calculations are selected. To find the most stable hydrogen adsorption site on these structures, 13 possible initial adsorption sites are optimized by DFT calculations. Figure 3a,b shows the most stable hydrogen binding site on single TM-doped g-C 3 N 4 , in which two potential adsorption sites for H are obtained. For g-C 3 N 4 doped with Ag, Co, Cu, Fe, or Ni, hydrogen atom bonds to the carbon atom near the doped TM. However, hydrogen binds with the nitrogen atom whenever the dopant is Cd, Pt, Sc, Ti, or Zn. In contrast to single TM-doped g-C 3 N 4 in which the H binding site is dependent on the doped TM, in TM-B co-doped structure the H atom always binds with C atom close to the doped TM. (Figure 3c). Figure 4 shows the energy profile of hydrogen adsorption on the surface of the single TM-doped and the TM-B co-doped g-C 3 N 4 monolayer. For single metal-doped g-C 3 N 4 , C─H, or N─H bond can form depending on the doped TM. By comparing the adsorption energies of hydrogen in the single metal-doped structures (Figure 4a,b), we can see that structures with N─H bond have lower adsorption energy than those with C─H bond. Bader charge analysis show that the nitrogen atoms in the g-C 3 N 4 monolayer structures are negatively charged, and carbon atoms are positively charged. It is generally known that in a structure, atoms with positive charge (here, C atoms) are more willing to obtain electrons than the atoms with negative charge (here, N atoms). As hydrogen has one electron and willing to donate the electron, carbon prefers to get the electron and binds with the hydrogen strongly in comparison with that of N. Therefore, the C─H bonds are expected to be stronger than N─H bonds. The energy profile shows a smaller adsorption energy for the structures with hydrogen binding with nitrogen atom (Pt, Zn, Cd, Ti, and Sc-doped g-C 3 N 4 ) and larger adsorption energy for structures with bonded hydrogen atom to carbon (Cu, Ag, Ni, Co, and Fe-doped g-C 3 N 4 ).
In all the co-doped structures, hydrogen is bonded to the carbon atom near the doped TM at edge of the hole. Considering the structures and energy levels (Figure 3 and 4), TM-doped structures can be divided into three categories. The first category is the single metal-doped structures in which hydrogen is bonded to nitrogen (Pt, Zn, Cd, Ti, and Sc-doped g-C 3 N 4 ). Introduction of boron as a co-dopant into these structures can change the hydrogen bonding site from nitrogen to carbon. Compared to N─H bonding in the single-doped structures, due to stronger C─H bonding in the co-doped structures, the adsorption energy shifts toward more negative values. This means that co-doping C 3 N 4 monolayer structure with boron and these metals can decrease the HER performance of these structures. Both the second and third category of the single metal-doped g-C 3 N 4 have C─H bonding. When the doped TM is Ni, Co, or Fe (second category), the hydrogen adsorption energies slightly shift toward more negative energies. Therefore, there is an increase in the adsorption energy of hydrogen in the second category structures. In Cu-B and Ag-B co-doped structures (the third category), however, the hydrogen adsorption energy shifts toward the zero. This upward shift means a reduction in adsorption energy of hydrogen on these structures. This category of structures (especially Cu-B co-doped structure) is promising candidates for HER. When the energy diagram of these samples are compared with the reported Pt-like structures with high HER efficiency, it seems that these structures are ideal candidates for HER. [29] To understand the difference between the hydrogen adsorption energies of the three categories, Bader charge analysis and charge density difference (CDD) are done. The Bader charge values show that for structures with N─H bonding, the Bader charge of hydrogen in N─H bond is more positive than that in C─H bond. (Tables S1 and S2 in Supporting Information) Nitrogen atoms in the structure of g-C 3 N 4 are electron acceptors (electronegativity of 3.04) and carbon atoms are electron donors (electronegativity of 2.55). Hydrogen atom (electronegativity of 2.2) has lower electronegativity than carbon, which explains why nitrogen prefers to get electrons from hydrogen instead of carbon. For the single-doped and co-doped structures with C─H bonding (TM-doped and TM-B co-doped, TM ¼ Cu, Ag, Ni, Co, and Fe), it seems that Ag and Cu behave differently from Ni, Co, and Fe. Strong hydrogen bonding exists on the Cu-and Ag-doped structures. Bader charge analysis shows that the hydrogen's electrons mainly flows to the TMs, and the TMs become less positive (Cu: from 0.83 to 0.76 and Ag: from 0.68 to 0.61). After hydrogen adsorption, unlike the Cu-and Ag-doped samples, the Bader charges of Ni (from 0.75 to 0.77), Co (from 0.86 to 0.88), and Fe (0.98 to 1.02) in the doped samples slightly increase. In Cu-doped and Ag-doped g-C 3 N 4 , the Bader charges of the bonded carbon change from 0.94 to 0.91and 1.07 to 1.03, respectively. These values show that there is a moderate change in the Bader charge of the bonded carbon. For the other samples that are bonded slightly weaker to H (Ni, Co, and Fe), however, electrons from hydrogen mainly flows to the carbon atom that is bonded to the hydrogen (see Table S1, Supporting Information). For Co-, Ni-, and Fe-doped systems, the carbon's charge changes from 0.95 to 0.84, 1.02 to 0.91, and 0.88 to 0.76, respectively. This means that electron from hydrogen goes mostly to carbon atom. Therefore, hydrogen bonding in these structures is softer than hydrogen bonding in the Cu-doped and Ag-doped g-C 3 N 4 structures. In Cu-and Ag-doped structures, there is significant charge transfer between Cu/Ag and the carbon atom, leading to a strong binding between carbon and H.
For the co-doped samples, the overall trend for Bader charge transfer to the TMs and bonded carbon (C─H) atoms are similar. After hydrogen adsorption, in all the Cu, Ag, Co, Ni, and Fedoped structures, Bader charge of the TMs increases, and the Bader charge of the bonded carbon atoms decreases. The reason for the weaker C─H binding in Cu-B and Ag-B co-doped structures is the doping-induced electronic structure change. Ag and Cu have 4d 10 , 5s 1 and 3d 10 , 4s 1 electronic structures, respectively. This means that in Ag and Cu atoms, s orbital is more important in donating/accepting electrons. For the rest of the considered www.advancedsciencenews.com www.small-structures.com TMs, however, the d electrons are more important. It is generally known that the 4s orbitals have lower energies than 3d orbitals. CDD analysis is also used to better understand the difference between these two categories with C─H bonding. In the singledoped structures, the hydrogen is bonded to a carbon atom which is bonded to the doped TM. The CDD analysis of the singlemetal-doped samples is shown in Figure 5a-d, indicating that the transferred charge from carbon to Ag and Cu is less than that transferred from the carbon atom to Ni and Co atoms. This result agrees well with the Bader charge analysis. In the co-doped structures, the TM is surrounded with nitrogen atoms. At least one of these nitrogen atoms is interacting with the bonded carbon (C─H) atom. As shown by the CDD results in Figure 5d-f, the substantial charge transfer between the nitrogen atom and Ni in Ni-B co-doped structure indicates the strong bonding for C─H. For the other co-doped structures, however, the charge transfer is less, and softer C─H bonds is expected. These results are also in good agreement with the Bader charge analysis.
The Pt-doped structure can give the high efficiency for the HER that is consistent with the literature. [30] Pt is an expensive metal, and thus not suitable for large-scale applications. Considering the cost of the dopants and the computational results from this study, Zn as a single dopant and Cu-B as a co-dopant are cheaper alternatives to improve the catalyst performance of g-C 3 N 4 monolayer. Among the considered metals, the Ag and Cu as single dopants in the structure of the g-C 3 N 4 monolayer show the worst efficiency for the HER. The introduction of boron to the structure of Pt-doped g-C 3 N 4 can alter the Pt-doping site and worsen the catalyst performance of the Pt-B co-doped g-C 3 N 4 structure.

Conclusion
In this study, DFT is used to investigate the doping site of ten TMs in the structure of g-C 3 N 4 monolayer. It is found that the dopants can break the C─N bond at edge of the hole, and thus deform the 2D structure of the g-C 3 N 4 . Among all the considered structures, Pt shows the best performance for HER, with a hydrogen adsorption energy of 0.1 eV. The second and third best structures for HER are Zn-doped and Cu-B co-doped g-C 3 N 4 structures with a hydrogen adsorption energy of 0.2 and 0.4 eV, respectively. It is found that the introduction of TMs as single dopants can deform the structure of g-C 3 N 4 monolayer. However, TM-B co-doping can maintain the 2D structure of g-C 3 N 4 compared with that of the single metal-doped system. Among all the structures, Ag and Cu has different doping behavior due to their unfilled s orbitals. Based on the computational results, Cu-B co-doped g-C 3 N 4 structure, because of its better 2D structural stability and low hydrogen adsorption energy, is an ideal material for HER.

Supporting Information
Supporting Information is available from the Wiley Online Library or from the author.