Assessment of Edge Modification of Nanographene Authors: Ryo Sekiya, Takeharu Haino Published: 2024 ABSTRACT AbstractCarboxy groups on the edges of nanographene (NG) enable functionalization for realizing NG‐organic hybrid materials. Therefore, assessment of the edge‐functionalization of the electronic structures of NGs is valuable for the rational design of functional carbon materials. In this study, the structures of model NGs comprising 174 carbon atoms with armchair edges and various functional groups at the edges were computed. To achieve the greatest possible similarity between the computed structure and the real one, the carbon framework was designed based on experimental observations. The functional groups can be accessed via suitable chemical reactions. The computations predicted that although the conversion of carboxyl groups with electron‐withdrawing/donating groups influences the orbital energies, the HOMO‐LUMO (H−L) gap is not significantly affected, except in a few cases. Among the evaluated examples, π‐extension had the greatest influence on the H−L gap. Interestingly, for the Pd2+‐coordinated NG, the participation of the low‐lying LUMO localized on Pd2+ in the surface‐to‐metal transitions seemingly narrowed the H−L gap, and a surface‐to‐ligand transition was observed. Introduction Structural modifications such as heteroatom doping or the installation of organic functional groups at the edges can perturb the electronic structures of graphitic materials. For example, the solvothermal reactions of three regioisomers of phenylene diamines yield nitrogen‐doped nanographenes (NGs), also known as graphene quantum dots,[ 1 , 2 , 3 ] displaying blue, red, and green emissions. [4] Similarly, citric acid and urea in formamide solution produce nitrogen‐doped NGs of various colors. [5] Density functional theory (DFT) calculations utilizing model NGs suggest that the nitrogen atoms and their positions in the carbon frameworks influence the HOMO‐LUMO (H−L) gap, leading to emissions with distinct colors. An example of edge functionalization is the post‐synthetic modification of NGs produced by top‐down methods. The oxidative cleavage of graphite with strong acids provides NGs with armchair edges as the dominant edge structures (Figure 1). [6] Carboxy groups on the edge [7] allow the installation of organic functional groups for realizing NG‐organic hybrid materials, such as those reported by our group[ 7 , 8 , 9 , 10 , 11 , 12 , 13 , 14 , 15 ] and Martín's group.[ 16 , 17 , 18 ] An intriguing finding is the drastic change in the optical spectra of NGs upon edge‐functionalization with naphthalene‐1,8‐diamine derivatives. [10] Computational analysis suggested that the electronic interactions between the NGs and the installed functional groups influence the energy levels of the NGs. [19] Diagram of NG (left), its edge structure (right), and edge functionalization (bottom). NGs produced by acid‐assisted oxidative cleavage of graphite have an armchair edge with two carboxy groups on the edge as dominant edge structures. These studies demonstrate that computational approaches are helpful for investigating and predicting the physical properties of NGs.[ 19 , 20 , 21 , 22 , 23 , 24 , 25 , 26 , 27 , 28 ] This background led us to study the effects of substituting carboxyl groups with various functional groups (alcohol, aldehyde, amine, bromine, aromatization, and metal coordination) on the electronic structures of NGs through computational analyses. The basal plane was designed based on the experimentally obtained NGs. [7] NMR and IR spectroscopic analyses demonstrated that the carboxy‐terminated armchair edge was the dominant edge structure of the NGs. These substituents are accessible via suitable chemical reactions on the carboxy‐terminated NGs. DFT calculations predicted that although the orbital energies depend on the electron‐withdrawing/donating nature of the substituents, the H−L gap is not significantly affected, except in a few cases. Furthermore, metal coordination can modify the electronic structure of NGs because of the electron‐withdrawing nature and the participation of the metal cations and ligands in the electronic transitions. Results and Discussion Figure 2 shows the model NGs (NG1–NG10) used in this work. The carbon framework of NG1 was designed based on experimental observations. [19] Primary alcohol (NG2)‐, aldehyde (NG3)‐, and amine (NG4)‐terminated NGs were studied to evaluate the effects of electron‐withdrawing and electron‐donating groups on the electronic structure of the carbon frameworks. These functional groups can be generated by reduction (alcohol) and mild oxidation of the resulting alcohol (aldehyde) using inorganic reducing and oxidizing reagents, respectively. The substitution of the carboxyl group with an amino group can be realized by Curtius or Hofmann rearrangement (amine). When the amination did not proceed quantitatively, the carboxyl and amino groups coexisted on the edge. This led to the design of partially aminated NGs (NG5 and NG6). These NGs are more realistic models for the experimentally obtained NGs. The amino groups can react with diketones to produce pyrazine rings. This inspired the design of NG7, which was obtained from NG6 through conversion of the amino groups to pyrazine rings. The amino groups of NG5 can be brominated via the Sandmeyer reaction; thus, bromine‐substituted NG (NG8) was designed. Bromine‐containing NGs are important targets for hybrid materials because the C−Br bond can be converted into a C−C bond via catalytic reactions. The amino groups of NG5 coordinate with the metal cations to form metal‐coordinated NGs. Positively charged metal cations influence the orbital energies owing to the electron‐withdrawing nature of the metal cations. Tetrahedral monocationic Cu+ (NG9) and dicationic square‐planar Pd2+ (NG10) complexes were prepared. In both cases, pyridine rings occupied the remaining coordination sites. Model NGs (NG1–NG10) used in this study. The surfaces of the NGs are drawn with single lines for clarity. Structural optimizations were performed using the Gaussian16 Rev. C01 program at the cam‐B3LYP level of theory with the 6–31G(d) basis set. [29] The coordinated metal cations on NG9 and NG10 were treated using the LanL2DZ method. Frequency calculations for the optimized structures were performed to confirm the absence of negative frequencies. The absorption spectra were calculated using the time‐dependent (TD)‐cam‐B3LYP/6‐31G(d) method. A total of 100 transitions was calculated. To check the dependence of the DFT methods on the energy levels and the trend of the energy change, ωB97XD and M06‐2X methods were also employed. The optimized structures of NG2–NG10 are shown in Figure 3, and NG1 [19] is shown for comparison. The edges of NG1 adopt alternating upward and downward structures. We previously reported that hydrogen‐terminated NG (i. e., non‐oxidized NGs) is predicted to have a flat edge. [28] Hence, the puckered edges were caused by the tendency of neighboring carboxy groups on the edges to avoid steric contacts. Similar puckered edges were predicted for NG2–NG10. Therefore, the edge structures of the experimentally obtained NGs were not expected to change drastically after substituting the carboxyl groups with other functional groups. Optimized structures of (a) NG1, (b) NG2, (c) NG3, (d) NG4, (e) NG5, (f) NG6, (g) NG7, (h) NG8, (i) NG9, and (j) NG10 at the cam‐B3LYP/6‐31G(d) level. The Cu+ and Pd2+ cations in NG9 and NG10 were treated by the LanL2DZ method. Color scheme: gray (carbon), white (hydrogen), blue (nitrogen), red (oxygen), orange (copper), purple (palladium). The dotted lines in (i) and (j) denote the coordination bonds. The planarity of the surface was evaluated using the least‐squares planarity index (LSPI), defined by Equation 1: (1) LSPI=1n∑inri where r i is the distance between carbon i and the least‐squares plane defined by the carbon framework, excluding the carbon atoms at the edge (Figure S1a). The LSPI of NG1–NG10 was <0.1 Å, with small standard deviations (Table 1). When all the carbon atoms of the framework were considered in the calculations (Figure S1b), the LSPI increased, with larger standard deviations (LSPI‐all in Table 1), demonstrating that the distortion was limited to the edge. Table 1LSPI (/Å) values for the optimized structures of NG1–NG8 at the cam‐B3LYP/6‐31G(d) level, those of NG9 and NG10 at the cam‐B3LYP/6‐31G(d)+LanL2DZ level, and the orbital energy levels (eV) of HOMO‐2 to LUMO+2 at the cam‐B3LYP/6‐31G(d) level. NG LSPI [Å] LSPI‐all [Å] HOMO−2 [eV] HOMO−1 [eV] HOMO [eV] LUMO [eV] LUMO+1 [eV] LUMO+2 [eV] H−L gap [eV] NG1 0.06(4) 0.25(30) −6.9756 −6.5101 −6.2937 −3.1671 −2.9617 −2.8327 3.1266 NG2 0.07(5) 0.25(32) −5.5922 −5.1772 −5.0249 −1.6814 −1.5110 −1.1184 3.3435 NG3 0.08(6) 0.24(28) −7.6834 −7.1765 −6.9819 −3.7995 −3.6910 −3.5497 3.1824 NG4 0.09(7) 0.24(29) −4.9993 −4.6221 −4.5171 −1.2253 −1.1633 −0.6095 3.2918 NG5 0.07(5) 0.24(28) −6.3253 −5.8205 −5.6325 −2.6324 −2.3976 −2.0191 3.0001 NG6 0.08(6) 0.25(29) −6.5873 −6.0352 −5.9310 −2.7905 −2.6591 −2.4351 3.1405 NG7 0.06(5) 0.22(28) −6.7862 −6.5107 −5.9895 −3.1059 −2.6700 −2.6183 2.8836 NG8 0.06(4) 0.23(29) −6.9082 −6.4657 −6.2788 −3.0240 −2.8229 −2.6784 3.2548 NG9 0.06(4) 0.22(27) −7.5721 −7.1751 −6.9939 −3.7560 −3.6082 −3.3100 3.2379 NG10 0.06(4) 0.22(27) −8.8132 −8.4252 −8.2391 −5.5968 −5.2276 −5.1576 2.6422 Wiley‐VCH GmbH LSPI (/Å) values for the optimized structures of NG1–NG8 at the cam‐B3LYP/6‐31G(d) level, those of NG9 and NG10 at the cam‐B3LYP/6‐31G(d)+LanL2DZ level, and the orbital energy levels (eV) of HOMO‐2 to LUMO+2 at the cam‐B3LYP/6‐31G(d) level. NG LSPI [Å] LSPI‐all [Å] HOMO−2 [eV] HOMO−1 [eV] HOMO [eV] LUMO [eV] LUMO+1 [eV] LUMO+2 [eV] H−L gap [eV] NG1 0.06(4) 0.25(30) −6.9756 −6.5101 −6.2937 −3.1671 −2.9617 −2.8327 3.1266 NG2 0.07(5) 0.25(32) −5.5922 −5.1772 −5.0249 −1.6814 −1.5110 −1.1184 3.3435 NG3 0.08(6) 0.24(28) −7.6834 −7.1765 −6.9819 −3.7995 −3.6910 −3.5497 3.1824 NG4 0.09(7) 0.24(29) −4.9993 −4.6221 −4.5171 −1.2253 −1.1633 −0.6095 3.2918 NG5 0.07(5) 0.24(28) −6.3253 −5.8205 −5.6325 −2.6324 −2.3976 −2.0191 3.0001 NG6 0.08(6) 0.25(29) −6.5873 −6.0352 −5.9310 −2.7905 −2.6591 −2.4351 3.1405 NG7 0.06(5) 0.22(28) −6.7862 −6.5107 −5.9895 −3.1059 −2.6700 −2.6183 2.8836 NG8 0.06(4) 0.23(29) −6.9082 −6.4657 −6.2788 −3.0240 −2.8229 −2.6784 3.2548 NG9 0.06(4) 0.22(27) −7.5721 −7.1751 −6.9939 −3.7560 −3.6082 −3.3100 3.2379 NG10 0.06(4) 0.22(27) −8.8132 −8.4252 −8.2391 −5.5968 −5.2276 −5.1576 2.6422 Table 1 summarizes the orbital energies (HOMO‐2 to LUMO+2) and H−L gaps of NG1–NG10, calculated using the cam‐B3LYP/6‐31G(d) method. The Cu+ and Pd2+ cations in NG9 and NG10 were treated using the LanL2DZ method. The corresponding energies evaluated by the M06‐2X and ωB97XD methods are compiled in Tables S2 and S3 in the Supporting Information. The change in the orbital energies of NG1–NG10 at the cam‐B3LYP/6‐31G(d) and ωB97XD/6‐31G(d) levels are shown in Figure 4a and b, respectively. Figure 5 summarizes the changes in the HOMO and LUMO energies relative to those of NG1 and H−L gap at the cam‐B3LYP/6‐31G(d) and ωB97XD/6‐31G(d) levels. The energy changes at the M06‐2X/6‐31G(d) level are shown in Figure S2 (Supporting Information). Changes in orbital energies of HOMO‐5 to LUMO+5 of NG1–NG10 at (a) the cam‐B3LYP/6‐31G(d) and (b) ωB97XD/6‐31G(d) levels. The changes in the orbital energies of HOMO‐5 to LUMO+5 of NG1–NG10 at M06‐2X/6‐31G(d) are shown in Figure S2 in the Supporting Information. (a) and (b) Difference between HOMO and LUMO gap of NG2–NG10 relative to that of NG1 at (a) cam‐B3LYP/6‐31G(d) and (b) ωB97XD/6‐31G(d) levels. (c) and (d) H−L gaps of NG1–NG10 at (c) cam‐B3LYP/6‐31G(d) and (d) ωB97XD/6‐31G(d) levels. DFT calculations predicted a dependence of the HOMO and LUMO energies on the substituents and their number. As expected, electron‐donating substituents (e. g., primary alcohols and amino groups) increased both the HOMO and LUMO energies, whereas electron‐withdrawing groups (aldehyde) lowered the orbital energies. The orbital energies gradually increased with an increase in the number of amino groups (NG6400 nm) were assigned as π‐π* transitions on the surface (Tables S4–S12 and Figures S3–S11 in the Supporting Information). The lowest‐energy transitions were predicted to occur between 631 and 670 nm, except in the case of NG7 and NG10. Figure 6a shows the electronic transitions of NG6 (red bars) and NG7 (black bars) calculated at the TD‐cam‐B3LYP/6‐31G(d) level of theory. A striking difference is the H−L transition of NG7 at 720.58 nm, which was significantly red‐shifted compared to that of NG6 at 659.52 nm, confirming that π‐extension is more effective for narrowing the H−L gap. (a) and (b) Calculated transitions (>500 nm) of (a) NG6 (red bars) and NG7 (black bars) and (b) NG5 (blue bars) and NG10 (black bars) at the TD‐cam‐B3LYP/6–31 G(d) level. (c) HOMO‐1 to LUMO+1 isosurfaces (isovalue=0.02) of NG10 at the cam‐B3LYP/6‐31G(d) level. The Pd2+ cations in (b) and (c) were treated by the LanL2DZ method. Color scheme: gray (carbon), white (hydrogen), blue (nitrogen), and red (oxygen). The narrower H−L gap of NG10 originates from the participation of the LUMO localized on Pd2+ in the electronic transitions. Figure 6c shows the HOMO‐1 to LUMO+1 iso‐surfaces of NG10 at the cam‐B3LYP/6‐31G(d) level. The HOMO and lower‐energy orbitals are on the surface, the LUMO is localized on Pd2+, and LUMO+1 is on the ligand. LUMO+2 is on another ligand, and LUMO+3 and higher‐energy orbitals are on the surface (Figure S11 in the Supporting Information). The low‐lying LUMO of Pd2+ narrows the H−L gap. TD‐DFT calculations at the TD‐cam‐B3LYP/6‐31G(d) level predicted that the surface‐to‐metal transition should occur at 730.82 nm (Figure 6b), despite having a very low oscillator strength (0.0022). The surface‐to‐ligand transitions were also predicted at 534.94 nm (HOMO‐>LUMO+1) and 516.90 nm (HOMO‐>LUMO+2). Low‐energy–π* transitions on the surface were predicted at 655.34 and 649.94 nm. The wavelength of these transitions is similar to those of NG5 (Figure 6b) and the other NGs. Hence, Pd2+ coordination does not significantly change the energy required for π‐π* transitions on the surface. Interestingly, the partially aminated NGs (NG5 and NG6) were predicted to have narrower H−L gaps (3.00 eV for NG5 and 3.14 eV for NG6) than the fully aminated congener (3.29 eV for NG4) at the cam‐B3LYP/6‐31G(d) level. The same trend was observed with the other DFT methods. The isosurfaces of the selected HOMO‐1 to LUMO+1 of NG5 and NG6 at the cam‐B3LYP/6‐31G(d) level are shown in Figure 7 (HOMO‐5 to LUMO+5 are shown in Figures S5–S7 in the Supporting Information). In both cases, HOMO‐1 and HOMO were localized near the amino‐terminated edge, and LUMO and LUMO+1 were localized near the carboxy‐terminated edge. TD‐DFT calculations confirmed the transitions from the HOMOs localized near the amino groups to the LUMOs near the carboxyl groups. (a) and (b) HOMO‐1 to LUMO+1 isosurfaces (iso‐value =0.02) of (a) NG5 and (b) NG6 at the cam‐B3LYP/6‐31G(d) level. Color scheme: gray (carbon), white (hydrogen), blue (nitrogen), and red (oxygen). The differences between the HOMO of the carbonyl‐terminated NG (NG1) and HOMOs of the fully aminated (NG4) and partially aminated NGs (NG6) (ΔHOMO) were −1.7766, −0.6612, and −0.3627 eV, respectively, and those between the LUMO of NG1 and that of NG4–NG6 (ΔLUMO) were −1.9418, −0.5347, and −0.3766 eV, respectively, at the cam‐B3LYP/6‐31G(d) level. In the case of NG4, the difference in the HOMO relative to the LUMO (ΔHOMO‐ΔLUMO; positive and negative values indicate widening and narrowing of the H−L gap, respectively, and nil indicates no change in the H−L gap) was +0.1652 eV, whereas ΔHOMO‐ΔLUMO for NG6 was +0.0139 eV. NG4 (3.2918 eV) had a wider H−L gap than NG1 (3.1266 eV), whereas NG6 (3.1405 eV) had a H−L gap similar to that of NG1. In the case of NG5, the amine‐terminated edge increased ΔHOMO (−0.6612 eV), while ΔLUMO remained at −0.5347 eV, and ΔHOMO‐ΔLUMO was −0.1265 eV. Hence, the H−L gap of NG5 was narrower than that of NG4 and NG6. These calculations suggest that partial amination or the coexistence of electron‐withdrawing and electron‐donating groups can narrow the H−L gap. Conclusions In conclusion, the substitution of the carboxy groups of NG, which are models of experimentally obtainable NGs, with electron‐withdrawing/donating groups can influence the orbital energies. The degree of change depends on the substituents and their number. Electron‐donating groups increase the HOMO and LUMO energy levels, whereas electron‐withdrawing groups decrease the orbital energy levels. Metal coordination significantly lowers the energy levels of the orbitals. Because the changes in the HOMOs and LUMOs are similar in degree, irrespective of the substituents, the H−L gap is not drastically changed by the substitution of the carboxyl groups with other functional groups, except in a few cases. Comparison of the calculated model NGs indicated that π extension is more effective in narrowing the H−L gap. Furthermore, the coordination of Pd2+ also influences the electronic structures of the NGs owing to the participation of the metal cations and ligands on the metal cations in the electronic transitions. Experimental Section Molecular modelling of all model NGs was performed using Winmostar ver. 10 program[30] on Windows 11Pro equipped with a Corei9 Extreme Processor with 128 GB memory. DFT calculations were performed using the Gaussian16 Rev.C01 program [29] on Linux workstations equipped with dual XEON silver 4316 processors with 256 GB ECC registered memory. Conflict of Interests The authors declare no conflict of interest.