Supplementary Materials for

Transcription

1 advances.sciencemag.org/cgi/content/full/4/4/e /dc1 Supplementary Materials for A physical catalyst for the electrolysis of nitrogen to ammonia Yang Song, Daniel Johnson, Rui Peng, Dale K. Hensley, Peter V. Bonnesen, Liangbo Liang, Jingsong Huang, Fengchang Yang, Fei Zhang, Rui Qiao, Arthur P. Baddorf, Timothy J. Tschaplinski, Nancy L. Engle, Marta C. Hatzell, Zili Wu, David A. Cullen, Harry M. Meyer III, Bobby G. Sumpter, Adam J. Rondinone This PDF file includes: Published 27 April 2018, Sci. Adv. 4, e (2018) DOI: /sciadv Modeling and simulation details table S1. Elemental analysis of C, N, O, and Si in original CNS, O-etched CNS, and glassy carbon by energy dispersive x-ray spectrometry elemental mapping. table S2. Partial current densities (ma cm 2 ) for CNS, oxygen-etched CNS, glassy carbon, and CNS with argon gas. table S3. Comparison of open-circuit potentials and polarizations of original CNS, O-etched CNS, and glassy carbon in 0.25 M KClO4. fig. S1. Representative TEM images of the CNS electrode. fig. S2. The variation of surface electric field Es calculated along the normal direction at the tip of a CNS for different tip radii in the case of desolvated Li + counterion and of solvated Li + counterion. fig. S3. Molecular dynamics simulation of electric double layers near a carbon nanosphere immersed in LiCl solution. fig. S4. Regression curves for ammonia quantification. fig. S5. SEM micrographs of CNS surface. fig. S6. SEM micrographs for the side view of CNS. fig. S7. XPS spectra of CNS. fig. S8. The overall current density (red curve) and formation rate (blue dots) with time at 1.19 V versus RHE. fig. S9. CP experiment to investigate stability of the electrode during the initial 5 hours of the reaction, using a larger (4.8 cm 2 ) electrode to observe changes with respect to electrolyte composition. fig. S10. Oxygen-etched CNS showing smoother texture compared to unetched CNS.

2 fig. S11. Correlated orbital levels calculated for three outer valence orbitals and three virtual orbitals at the level of EPT/aug-cc-pVTZ as a function of electric field strength. fig. S12. Ultravoilet photoelectron spectroscopy and work functions of emersed CNS. fig. S13. Ultraviolet photoelectron spectroscopy of pristine, unemersed CNS. fig. S14. Ultraviolet photoelectron spectroscopy of O-etched CNS and glassy carbon. fig. S15. Mass spectra of double-silylated product for ammonia from electrochemical N2 reduction. fig. S16. Mass spectra of 14 N and 15 N products in the mass region of the molecular ion. References (44 54)

3 Modeling and simulation details Simulation of the electrical field near a carbon nanospike The distribution of the electric field at the tip of a carbon nanospike (CNS) was estimated by using an exohedral electric double-sphere capacitor (xedsc) model (17). The strongly curved surface on the sharp tip was simulated by an electrically charged nanosphere, where counterions are electrostatically accumulated to form an xedsc. The capacitance of xedsc is given by C/A = ε r ε 0 (a + d)/ad (S1) where C is the capacitance, A is the specific surface area, r is the electrolyte dielectric constant, 0 is the permittivity of vacuum, a is the radius of the tip, and d is the double layer thickness. When a voltage V is imposed between the nanosphere and bulk electrolyte, the electric field on the surface of the nanosphere is then given by E s = V(a + d)/ad (S2) It follows that the enhancement of the surface electric field Es at the strongly curved surface is closely related to the imposed voltage V, the surface curvature of the nanosphere as characterized by its radius a, and the thickness of the double layer as measured with d. First of all, the imposed voltage V is determined by the difference between the polarized electrode potential of CNS and the potential of the bulk electrolyte. When the electrode was polarized to V vs. RHE as shown in Fig. 2A, the formation rate of ammonia is the highest. In order to obtain the imposed voltage V, the potential of the bulk electrolyte needs to be known. On the basis of the halfdifference of the redox potentials of water, the position of the Fermi level in neutral water is determined to be 0.62 V vs. RHE (44). Therefore, the potential difference between the polarized

4 CNS electrode and the bulk electrolyte is obtained as 1.81 V. It is reasonable to assume that this potential difference fully falls in the electric double layer, giving the imposed voltage V = 1.81 V that enters eqn. S2 for the calculations. Herein we did not consider the effect of the supporting electrolyte LiClO4 because it should not significantly affect the Fermi level of water. Secondly, the double layer thickness d is determined by the size of the counterion Li + and whether it is solvated by water molecules. Based on our atomistic simulations shown below, the double layer thickness d is found to be 0.20 and 0.36 nm in the case of desolvated and solvated Li + counterion, respectively. The electric field at the surface of an electrically charged nanosphere Es is then calculated using eqn. S2 as a function of the nanosphere radius a. The results in fig. S2 clearly indicate that in both cases of desolvated and solvated Li + counterion, the electric field at the tip of a CNS is strongly dependent on the tip size. For a tip with a radius of 1 nm and a voltage drop of 1.81 V, the electric field on its surface enhances by nearly 2 V/nm compared to that on a planar electrode surface. For smaller tip radius, the electric field at the sharp tips can be enhanced to well above 10 V/nm (or > 1 V/Å) in case of desolvated counterion. The surface electric field levels off only for tips of 10 nm in radius where it approaches asymptotically toward that of a planar electrode surface. The rationalization of these results is straightforward by using eqn. S2. Conventionally, the electric field on a planar electrode surface is simply V/d. In comparison, the electric field on the strongly curved surface of a CNS simulated by a charged nanosphere differs from that of the planar surface by a coefficient of (a+d)/a, which can be greater than 1 for small radius and approach 1 for large radius. As the tip radius approaches the value of the double layer thickness, the coefficient

5 approaches 2, resulting in a doubling of the electric field. While the CNS tips do not universally reach that scale, the general effect can still be clearly recognized. Molecular dynamics (MD) simulations of the electric double layer Molecular system: To simulate the electric double layers at the tip of a CNS, we adopt a simulation system as shown in fig. S3A,B. A carbon nanosphere with a radius of 1.0 nm is used to mimic the sharp tip of a CNS. The carbon nanosphere is placed at the center of a simulation box measuring nm 3. The simulation box is periodic in all directions and filled with aqueous electrolyte. In our experiments, Li + ions are the optimal counterion for N2 reduction while ClO4 - ions are the co-ions in the electric double layer. However, since ClO4 - ions have very little effect on the counterion distribution near the heavily charged tip of the CNS, Li + and Cl - ions are herein taken as the counterions and co-ions in the MD simulations for simplicity. The numbers of Li + and Cl - ions inside the MD system are tuned so that their concentration at position far away from the carbon nanosphere s center is ~0.3 M, similar to that in the experiments, and so that the MD system is overall electrically neutral. The capacitance of nitrogen-doped graphite with small curvature is experimentally shown to be 0.22 F/m 2 (45 Zhang). Assuming a double layer thickness of d = 0.2 nm (see below for justification), the capacitance at the tip of a CNS with a radius of a = 1.0 nm is found to be 0.26 F/m 2 using eqn. S1. In our experiments, the voltage drop between CNS and bulk electrolyte is 1.81 V (see above). Therefore, the surface charge density at the tip of a CNS is estimated to be 0.47 C/m 2. Accordingly, small partial charges are decorated on the atoms of the carbon nanosphere so that the surface charge density of the carbon nanosphere is equal to 0.47 C/m 2.

6 Molecular model: Water molecules are modeled using the SPC/E model. Li + and Cl - ions are modeled as charged Lennard-Jones spheres with force fields taken from the literature (46). The surface of the carbon nanosphere is modeled using carbon atoms arranged into hexagonal lattice and the spacing between neighboring atoms is 0.14 nm, similar to that found in graphene sheets. The Lennard-Jones parameters for the carbon atoms are taken from the literature (47). The Lennard-Jones parameters between different atoms are obtained using the Lorentz-Berthelot combination rule. Simulation method: Simulations are performed using the NVT ensemble (T = 300 K). A cutoff radius of 1.3 nm is used to compute the Lennard-Jones potential. The electrostatic interactions are computed using the Particle Mesh Ewald (PME) method. An FFT grid spacing of 0.11 nm and a cubic interpolation for charge distribution are chosen to compute the electrostatic interactions in the reciprocal space. All carbon atoms are fixed during the simulation. The LINCS algorithm (48) is used to maintain the water geometry specified by the SPC/E model. Starting from a random configuration, the system is simulated for 5.0 ns to reach equilibrium. A production run of 15 ns is then performed to gather the statistics of various quantities, e.g., water and ion densities near the carbon nanosphere. All simulations are performed using the Gromacs code (49). High-level Electron Propagator Theory (EPT) Calculations Previously nonempirical calculations were performed for N2 aligned in a longitudinal electric field, showing that the energy levels of inner valence 2 g and 2 u orbitals (i.e. HOMO-4 and HOMO-3) decrease linearly with increasing field strength. Similar straight lines were also obtained for the remaining levels of the N2 molecule without showing actual figures (18). Herein we carried out EPT calculations (50, 51) using Gaussian 09 suite of program (52) to confirm this

7 observation and to explicitly show the energy levels of outer valence and also virtual orbitals as a function of field strength. N2 molecular structure was first optimized using density function theory B3LYP/6-31G* in the presence of a longitudinal or a transversal electric field. Then the correlated ionization potentials (IPs) for the three outer valence HOMO-2 through HOMO orbitals and the correlated electron affinities (EAs) for the three virtual LUMO through LUMO+2 orbitals were calculated at the level of EPT/aug-cc-pVTZ using the N2 molecular structures optimized in the corresponding electric field. The strength of electric field was set in a range from 0 to 0.09 a.u., corresponding to 0 to V/nm. The energy levels for the outer valence and virtual orbitals are shown in fig. S10 as a function of the electric field strength. As can be seen from fig. S10, the IPs (negative of orbital energies) for the occupied HOMO-2 through HOMO are nearly independent of field strength. In addition, the magnitudes of the IPs are all greater than 15 ev, indicating that it is energetically expensive to oxidize the rather stable N2 molecule. The smallest IP is , in excellent agreement with experimental value of ev (53). In comparison, the EAs (negative of orbital energies) for the unoccupied LUMO through LUMO+2 are sensitive to the field strength, esp. when the molecule is aligned in parallel with the electric field instead of perpendicular to the electric field. In the absence of an electric field, the smallest EA is ev, in good agreement with experimental value of -2.2 ev (54). The negative sign indicates that it costs energy to inject (or attach) an extra electron to the LUMO orbital of N2. Conversely, N2 - anion would undergo electron detachment automatically. These results are well in line with the fact that N2 is inert to both oxidation and reduction. However, in the presence of an external electric field, the EAs may change to a positive sign under a critical electric field, which indicates that it becomes energetically

8 favorable to reduce N2 by injecting electrons into the antibonding orbitals of N2 under strong applied electric field. The sharp spikes present on the CNS surface are expected to create the strong electric field necessary to reduce N2, which would otherwise be inert under normal condition.

9 Supporting Tables table S1. Elemental analysis of C, N, O, and Si in original CNS, O-etched CNS, and glassy carbon by energy dispersive x-ray spectrometry elemental mapping. Carbon Materials Element distribution (mass %) C N O Si Original CNS O-etched CNS Glassy Carbon

10 table S2. Partial current densities (ma cm 2 ) for CNS, oxygen-etched CNS, glassy carbon, and CNS with argon gas. Significant ammonia production only occurs with CNS. Oxygenetched CNS has N-doping but less texture than CNS. Glassy carbon contains neither texture nor N-doping. CNS in argon is pristine but lacks N2 reagent for electroreduction to ammonia. CNS E/V vs RHE Average Std. Dev O-etched CNS E/V vs RHE Average Std. Dev Glassy Carbon E/V vs RHE Average Std. Dev Argon with CNS Control E/V vs RHE

11 table S3. Comparisons of open-circuit potentials and polarizations of original CNS, O- etched CNS, and glassy carbon in 0.25 M KClO4. Carbon materials Open-circuit potential (V) a Polarization (V) b Original CNS O-etched CNS Glassy carbon a Electrode potential vs. Ag/AgCl. b Difference between the electrode s open-circuit potential o and the polarized potential p, where the polarized potential p is V vs. Ag/AgCl while the open-circuit potential o depends on the materials.

12 Supporting Figures fig. S1. Representative TEM images of the CNS electrode. The inset shows an HR-TEM image of an individual spike of CNS. Reproduced with permission from Ref (15).

13 Electric field E s (V/nm) Li + (desolv) Li + (solv) Tip radius of carbon nanospike (nm) fig. S2. The variation of surface electric field Es calculated along the normal direction at the tip of a CNS for different tip radii in the case of desolvated Li + counterion and of solvated Li + counterion. The double layer thickness is 0.20 and 0.36 nm for the desolvated and solvated Li + counterion, respectively. The horizontal dashed lines indicate the electric fields on planar surface electrode for the desolvated and solvated Li + counterion, respectively. fig. S3. Molecular dynamics simulation of electric double layers near a carbon nanosphere immersed in LiCl solution. (A-B) Snapshot of the MD system and zoom-in view of the distribution of ions and water molecules near the carbon nanosphere s surface. (C) Distribution of water molecules, Li + ions, and Cl - ions as a function of the radial distance from the surface of the carbon nanosphere obtained from MD simulations.

14 A fig. S4. Regression curves for ammonia quantification. Regression curves were obtained from standard solutions in 0 to 50 μm (A) and 0 to 500 μm (B) concentration range. The slope of the curve is μm 1 and μm 1, respectively.

15 A fig. S5. SEM micrographs of CNS surface. Comparing the SEM micrograph of CNS surface before (A) and after (B) a 6-hour experiment at 1.19 V versus RHE, no significant change in the texture of CNS was observed.

16 A B fig. S6. SEM micrographs for the side view of CNS. SEM micrographs for the side view of CNS were taken before (A) and after (B) a 6-hour experiment at 1.19 V versus RHE for N2 electrochemical reduction. The thickness of the CNS layer did not change after the reaction, indicating stability of the CNS under the reaction conditions.

17 A A CNS as prepared B CNS - after reaction fig. S7. XPS spectra of CNS. XPS spectra were collected using a Thermo Scientific K-Alpha XPS operating at a base pressure of mbar. XPS N1s spectra of CNS before (A) and after (B) a 6-hour experiment of N2 electrochemical reduction. (1) = pyridinic nitrogen, (2) = amine-like nitrogen, (3) = graphitic-like nitrogen, or pyrrolidine, pyrrole or pyridine nitrogen, which has similar binding energy.

18 fig. S8. The overall current density (red curve) and formation rate (blue dots) with time at 1.19 V versus RHE. The formation rate of ammonia is generally stable after an initial induction period.

19 fig. S9. CP experiment to investigate stability of the electrode during the initial 5 hours of the reaction, using a larger (4.8 cm 2 ) electrode to observe changes with respect to electrolyte composition. Wettability of the electrode increases initially with subtle oxidation. Eventually, the ph of the solution begins to rise, which impacts total current density through suppressed H2 evolution. NH3 production varies but is sustained. The production rate for this experiment is lower than for fig. S8 because of a different cell geometry to accommodate the larger electrode, and commensurate lower mass transport.

20 fig. S10. Oxygen-etched CNS showing smoother texture compared to unetched CNS. Energy dispersive spectroscopy indicates that the N-doping percentage is unchanged after etching.

21 fig. S11. Correlated orbital levels calculated for three outer valence orbitals and three virtual orbitals at the level of EPT/aug-cc-pVTZ as a function of electric field strength. (A) N2 is parallel to a longitudinal electric field. (B) N2 is perpendicular to a transversal electric field. Electron affinities for the occupied orbitals and ionization potentials for the unoccupied orbitals are both negative of the orbital energies.

22 A B fig. S12. Ultravoilet photoelectron spectroscopy and work functions of emersed CNS. (A) Example ultraviolet photoemission spectra for pristine CNS emersed at 1.9 V and (B) work functions of emersed CNS as a function of emersion potential. The photoemission spectra are complex and some samples exhibit a low energy shoulder that may be due to emission from the nanospike tips, while the higher energy shoulder may be due to emission from surrounding carbon atoms. All samples are from the same region of the same wafer.

23 fig. S13. Ultraviolet photoelectron spectroscopy of pristine, unemersed CNS. A large, lowenergy peak obscures the low-energy cutoff, and does not shift under sample biasing up to -4.5 V during the photoemission measurement. This peak disappears upon O-etching of the CNS, and can therefore be assigned to emission from the sharp nanospike tips.

24 A B fig. S14. Ultraviolet photoelectron spectroscopy of O-etched CNS and glassy carbon. Ultraviolet photoemission spectra for O-etched CNS (A) and glassy carbon control (B). The blunted tips emit much more like a flat surface, but with still significantly lower WF compared to glassy carbon.

25 Counts/A.U. A Si Si O O N H 14 N product. MW= m/z

26 Counts/A.U. B Si Si O O 15 N H 15 N product. MW= m/z fig. S15. Mass spectra of double-silylated product for ammonia from electrochemical N2 reduction. The mass spectra of double-silylated product for natural 14 N (A) and 15 N-labelled (B) ammonia show predominant shifted peaks at 497/498 m/z, and minor fragment peaks at 182/183 and 240/241 m/z. The high isotopic purity of the 15 N labelled sample indicates that the entirety of the product was from 15 N2 gas, and not from N dopant liberated from the CNS electrode.

27 Counts/A.U. 14 N product 15 N product 14 N molecular ion 15 N molecular ion m/z fig. S16. Mass spectra of 14 N and 15 N products in the mass region of the molecular ion.