Mechanisms of Hydrogen-Assisted CO2 Reduction on Nickel - Journal

Mar 21, 2017 - Fahdzi Muttaqien , Hiroyuki Oshima , Yuji Hamamoto , Kouji Inagaki , Ikutaro Hamada , Yoshitada Morikawa. Chemical Communications 2017 ...
0 downloads 0 Views 1MB Size
Communication pubs.acs.org/JACS

Mechanisms of Hydrogen-Assisted CO2 Reduction on Nickel Wei Lin, Kelsey M. Stocker,† and George C. Schatz* Department of Chemistry, Northwestern University, Evanston, Illinois 60208-3113, United States S Supporting Information *

species is produced.5,6,10−21 Vesselli and co-workers have performed a series of pioneering investigations regarding the CO2 reactivity on Ni(110).6,10−14 Recently, they derived, on the basis of experimental observations at near-ambient pressure and 300−525 K temperature, that the metastable activated CO2− on Ni(110) acts as the precursor for dissociation into CO and O in the RWGS reaction and as a reactive species in the Sabatier methanation process.6 They also performed ultrahigh vacuum (UHV) low temperature experiments and DFT calculations, which indicated that CO + OH is generated from CO2 hydrogenation via the ER mechanism with HOCO as a “hot” intermediate as the DFT results show the dissociation of HOCO is relatively easy (0.13 eV).14 Although this suggests that both redox and associative mechanisms can occur, no direct molecular dynamics simulation evidence has been given to illustrate molecular level details of the mechanisms. In the present work, we present the first realtime Born−Oppenheimer molecular dynamics (BOMD) simulation aimed at unravelling the CO2 hydrogenation reaction mechanisms under UHV low temperature conditions, for both thermal (7.8 meV) and high energy (0.6 eV) H atom velocities, providing new insights to the current understanding of this process.5,6,14 Although this study is concerned with atomic hydrogen as the reagent, our results are also relevant to the RWGS reaction that involves H2 as reagent, as H2 undergoes dissociative chemisorption with unit probability.22 First, we performed BOMD simulations of H impinging on CO2 adsorbed Ni(110) surface using the VASP software package.23,24 The details of the computational methods are described in the SI. Figure 1c shows the distribution of outcomes for H impinging on 2CO2@Ni(110) at 90 K. Overall, three-fourths of the incident H are reflected back to vacuum (green crosses in Figure 1c), whereas 24% of them reach the surface area highlighted in Figure 1c. Because Ekin(H) is only 90 K (7.8 meV), when H is within the van der Waals radius of the CO2 (Figure 1b and S1), it almost always reflects back to vacuum. When H impinges in the highlighted area of Figure 1c, H typically reaches the Ni surface, where adsorption on Ni(110) is a barrierless process with Eads = −2.7 eV.6,20 Because of its light mass, the adsorbed H gains a high translational energy from its initial adsorption on the surface. This “hot” and highly reactive atom does not fly into vacuum but instead undergoes a high kinetic energy diffusion. Because the barriers for H diffusing on Ni(110) are much lower than the translational energy obtained from H adsorption,20 the adsorbed H has the capability to migrate on the zigzag Ni(110) surface before it

ABSTRACT: Mechanistic details of catalytic reactions are critical to the development of improved catalysts. Here, we perform high quality Born−Oppenheimer molecular dynamics simulations of the reaction mechanisms associated with hydrogen-assisted CO2 reduction on Ni(110). The simulation results show direct theoretical evidence for both associative and redox mechanisms in the reaction of atomic hydrogen with CO2. Because H2 is dissociatively chemisorbed on Ni(110) with nearly unit probability, the mechanisms we find are also relevant to the reverse water-gas shift reaction (H2 with adsorbed CO2). Furthermore, we provide the first real-time demonstration of both Eley−Rideal (ER) and hot atom (HA) mechanisms when H impinges on adsorbed CO2, and we show that both occur even for low kinetic energies. The trade-off between ER or HA mechanisms is found to be strongly dependent on CO2 coverage. The results are compared with recent gas/surface measurements.

R

ecently, CO2 capture and reduction have been extensively investigated in the literature.1,2 Among the various approaches, thermocatalysis and electrocatalysis utilizing heterogeneous catalysts on solid−gas or solid−liquid interfaces to produce useful chemicals from CO2 have become active research topics.3,4 For the metal catalysts, Ni exhibits promising catalytic activity, either for dissociating CO2 to CO via the reverse water−gas shift (RWGS) reaction, or by Sabatier methanation, as reported in recent investigations on both Ni(111) and Ni(110).5,6 In general, the RWGS reaction can proceed via associative or redox mechanisms. In the associative mechanism, CO2 reacts with H to form carboxyl (HOCO) as an active intermediate, which splits into CO and OH followed by the hydrogenation of OH to H2O, whereas in the redox mechanism, CO2 first dissociates to CO and O followed by two steps of O hydrogenation to water. For either the associative or the redox mechanism, CO2 hydrogenation on the surface can occur through Eley−Rideal (ER), Langmuir−Hinshelwood (LH), or hot atom (HA) mechanisms, depending on how the reaction happens: either H directly attacks from the gas phase (ER), or H reacts after thermalization (LH), or H traps briefly on the Ni surface, becomes translationally hot as the >2 eV adsorption energy is released, and then impacts and reacts with CO2 (HA).7−9 Several experimental studies of CO2 hydrogenation mechanisms on Ni surfaces under different conditions have been reported, and there is evidence for the redox mechanism (CO2− dissociation to CO + O), the associative mechanism (HOCO intermediate), and a third mechanism in which a stable HCOO © XXXX American Chemical Society

Received: February 14, 2017 Published: March 21, 2017 A

DOI: 10.1021/jacs.7b01538 J. Am. Chem. Soc. XXXX, XXX, XXX−XXX

Communication

Journal of the American Chemical Society

impinging “hot” H and CO2. Here we note that before H atom impact, the chemisorbed CO2 is negatively charged (−0.9 e) on Ni(110) with C positive and both O’s negative. Of the four trajectories forming CO* + OH*, one trajectory follows the ER channel and the rest are via the HA channel (Table 1). In the ER channel (Figure 2a), the H reacts with Table 1. Number of Trajectories Following the Associative or Redox Mechanism with the ER or HA Channel at Different Initial H Kinetic Energies, Ekin(H) 7.8 meV (90 K)

600 meV

Ekin(H)

associative

associative

redox

ER HA

1 3

4 9

6 5

CO2 directly from the gas phase without adsorption. At 200 fs after the beginning of this trajectory, the H atom slows down as it surmounts the barrier adding to CO2, and then it vibrates rapidly in the HOCO* complex. In this complex, there is significant charge transfer from H to C, and at 350 fs in Figure 2a, transfer of OH vibrational energy to CO causes dissociation of the CO bond (∼150 fs after the H addition), forming adsorbed CO* and OH* with both carrying −0.45 e of charge. The time-dependent reaction process is shown in the Supporting Movie denoted ER. Three trajectories produce CO* + OH* following the HA channel (labeled as small red squares in Figure 1c). The bond distances and charges of the CO2 and H of one trajectory (initial position of H at x = 2.67 Å, y = 3.25 Å in Figure 1c) are shown in Figure 2b. In this HA channel (HA1), the H atom is first adsorbed on the surface, releasing significant energy. Subsequently, it bounces a couple of times (nonthermal diffusion) on Ni atoms that are in the crevice between the two CO2’s, and then abstracts O from a CO2 to form OH* at ∼240 fs in Figure 2b. (see Supporting Movie HA1.) Note in Figure 2b the rapid transfer of charge from H to C while the OH* is forming. Interestingly, the HA2 trajectory (x = 2.74 Å, y = 3.29 Å in Figure 1c) is similar to HA1 but leads to a longerlived HOCO* complex (900 fs lifetime in Figure 2c) before dissociation of the CO bond (see Supporting Movie HA2). Again, there is rapid transfer of charge from H to C when the OH bond forms, showing the importance of charge transfer in the reaction mechanism.6 The third HA trajectory (x = 3.09 Å, y = 3.71 Å in Figure 1c) is similar to the HA2 for the first ps (see Figure S6), giving HOCO* in which OH is vibrationally excited. However, at ∼1.2 ps, the H transfers between two CO2’s, causing dissociation of the C−O bond in the second CO2. Half a picosecond after the proton transfer, the first CO2 is desorbed to vacuum. Overall, we have provided evidence for both ER and HA channels during H impact on Ni(110) with an initial kinetic energy of 90 K. The simulation results confirm the experimental observation of H-assisted C−O bond cleavage, with HOCO* as an intermediate. Only the ER mechanism was assumed to occur in the experiments,14 so our direct simulation has provided previously unsuspected detail concerning the reaction mechanisms for UHV low temperature conditions. All four trajectories follow the associative mechanism, which indicates that the redox mechanism is not important at low temperature. Two trajectories associated with the HA channel generate HOCO* (see Figure 1c) that does not dissociate during the 3

Figure 1. (a) Side view of the system (H, gray; O, red; C, cyan; Ni, blue; unit cell, green dashed rectangle). (b) Top view, showing van der Waals radii for CO2 and H. (c) Distribution of outcomes for H impacts on the Ni(110) unit cell with Ekin(H) = 90 K. For the CO* + OH*, large or small red squares represent trajectories following the ER or HA mechanism.

either reacts or is thermalized. Overall, of the 48 nonreflected trajectories, H thermalizes on the surface without reacting 17 times, and it penetrates to the subsurface and eventually thermalizes on the surface 9 times. In three trajectories, H thermalizes in the bulk. The “hot” H also causes CO2 desorption in 8 trajectories. Representative trajectories showing reflected H, surface H, subsurface H, bulk H, and CO2 desorption are presented in Figures S1−S5. Four trajectories are found to form CO* + OH*. Using bond distances to show the evolution of the adsorbed species, important details of the different reaction mechanisms are revealed in Figure 2. In addition, a Bader charge analysis has been used in Figure 2 to describe charge transfer between the

Figure 2. Bond distances (top) and atomic Bader charges (bottom) of trajectories following the (a) ER and (b and c) HA mechanisms with Ekin(H) = 90 K. B

DOI: 10.1021/jacs.7b01538 J. Am. Chem. Soc. XXXX, XXX, XXX−XXX

Communication

Journal of the American Chemical Society

reactions happen when the initial H hits close to CO2 (large red squares or circles in Figure 3), while the HA reactions occur when H hits the Ni surface. Hence, the coverage of the CO2 on Ni(110) controls the overall ratio of ER and HA reactions. These two crucial points clarify the long-standing debate about the actual reaction path of CO2 hydrogenation on Ni.14 Furthermore, HCO3* is observed in one trajectory, where H attacks CO 2 through the ER channel and causes CO2 dissociation to give CO* and OH*, followed by OH* diffusion to react with a second CO2 to give HCO3* in a “reversed Y” configuration. The populations of products from two sets of simulations are displayed in Figure 4. The most probable fate of H at Ekin = 90

ps duration considered (and we extended these trajectories to 10 ps with no change in outcome). As displayed in Figure S7, for these trajectories, the H initially interacts with Ni for ∼200 fs before forming HOCO*. Partial thermalization occurs during this interval, leading to HOCO* that is less energetic, and there is further relaxation of the OH bond in HOCO* after 4 ps. As a result, the ∼0.4 eV barrier for HOCO* dissociation is not overcome during the 10 ps interval.20 Figure 1c also shows five trajectories that produce HCOO* by the HA mechanism. Here H forms the H−C bond after a few bounces on the Ni surface (HA mechanism). The H−C then rotates 90°, breaking the Ni−C bond, and the HCOO* then forms a “reversed Y” configuration (Figure S8), as reported in previous studies.12,14 The simulated HCOO* configuration confirms the DFT calculations shown in Figure 1 of ref 14, where HCOO* is energetically stable with the “reversed Y” shape. To consider a higher H atom translational energy (part of the broad distribution of energies sampled in the atomic hydrogen beam experiment14), we next ran 100 more trajectories with the same setting as the first 200 trajectories except Ekin(H) was increased to 0.6 eV. The distribution of outcomes of these 100 trajectories is shown in Figure 3. In contrast to the results in

Figure 4. Populations of products for H impact on Ni(110) with the Ekin(H) = 90 K and 0.6 eV.

K (7.8 meV) is reflection back to vacuum, whereas at Ekin(H) = 0.6 eV energy transfer from the incident H atom to CO2 leads to CO2 desorption. We also find that 10% of the trajectories with Ekin(H) = 0.6 eV involve a combination of desorption of one CO2 and reaction with the other. Also, the reaction fraction (sum of the populations in right five sets of columns in Figure 4) increases to 32% compared to 5.5% for Ekin(H) = 90 K (7.8 meV). Hence, our comparative studies indicate that the reactivity of the H-assisted CO2 reduction is significantly increased when the H has enough Ekin to react directly with the adsorbed CO2. The adsorption energies and reaction barriers of each mechanism in this study are listed in the Supporting Information. These results show that the barrier following the HA channel in the associative mechanism (1.1 eV) is much higher than in the redox mechanism (0.4 eV), whereas the corresponding ratio of reaction by these mechanisms is 9:5 at Ekin(H) = 0.6 eV (Table 1). The relative unimportance of the associative mechanism arises because there is competition between H-induced CO2 desorption and dissociation (redox mechanism), both of which involve ∼0.4 eV barriers. Also, in the HA mechanism, the H atom becomes translationally hot when the adsorption energy is released, so it is much easier for H to overcome the 1.1 eV barrier to association than for C−O bond breakage via the redox mechanism.25 In this work, we performed two sets of BOMD simulations to demonstrate H-assisted CO2 reduction on Ni(110) and we used the results to determine the reaction mechanisms. The results demonstrate that (i) H impinging on CO2 adsorbed

Figure 3. Distribution of outcomes for H impacts on Ni(110) with Ekin(H) = 0.6 eV. Large or small red squares (and open red circles) represent trajectories following the ER or HA mechanisms to give CO* + OH* (and CO* + O* + H*).

Figure 1, only 10% of the trajectories are reflected back to vacuum, whereas the reactive fraction (red symbols in Figure 3) is 32%, of which 13 trajectories follow the associative mechanism and 11 follow the redox mechanism. In detail, four and nine of the trajectories that involve the associative mechanism are via ER and HA pathways, respectively, as labeled by the large and small red squares in Figure 3 and listed in Table 1. Similarly, six and five of the redox mechanism trajectories follow the ER and HA channels, respectively. These results illustrate two key points regarding the mechanisms of Hassisted CO2 reduction on the Ni surface. First, we provide direct theoretical evidence for both associative and redox mechanisms for the reaction of atomic hydrogen with CO2 on Ni, which is consistent with HREEL and XPS spectra measurements.14 Second, the simulations show the existence of both ER and HA mechanisms for both Ekin(H) = 90 K and 0.6 eV. Our simulation results thus generalize the experimental conclusions that only the ER reaction arises from H impacts on CO2 adsorbed Ni(110).14 In detail, as listed in Figure 3, ER C

DOI: 10.1021/jacs.7b01538 J. Am. Chem. Soc. XXXX, XXX, XXX−XXX

Communication

Journal of the American Chemical Society

(11) Monachino, E.; Greiner, M.; Knop-Gericke, A.; Schlogl, R.; Dri, C.; Vesselli, E.; Comelli, G. J. Phys. Chem. Lett. 2014, 5, 1929−1934. (12) Vesselli, E.; De Rogatis, L.; Ding, X. L.; Baraldi, A.; Savio, L.; Vattuone, L.; Rocca, M.; Fornasiero, P.; Peressi, M.; Baldereschi, A. J. Am. Chem. Soc. 2008, 130, 11417−11422. (13) Vesselli, E.; Schweicher, J.; Bundhoo, A.; Frennet, A.; Kruse, N. J. Phys. Chem. C 2011, 115, 1255−1260. (14) Vesselli, E.; Rizzi, M.; De Rogatis, L.; Ding, X. L.; Baraldi, A.; Comelli, G.; Savio, L.; Vattuone, L.; Rocca, M.; Fornasiero, P. J. Phys. Chem. Lett. 2010, 1, 402−406. (15) Remediakis, I. N.; Abild-Pedersen, F.; Norskov, J. K. J. Phys. Chem. B 2004, 108, 14535−14540. (16) Studt, F.; Abild-Pedersen, F.; Wu, Q. X.; Jensen, A. D.; Temel, B.; Grunwaldt, J. D.; Norskov, J. K. J. Catal. 2012, 293, 51−60. (17) Peng, G. W.; Sibener, S. J.; Schatz, G. C.; Ceyer, S. T.; Mavrikakis, M. J. Phys. Chem. C 2012, 116, 3001−3006. (18) Peng, G. W.; Sibener, S. J.; Schatz, G. C.; Mavrikakis, M. Surf. Sci. 2012, 606, 1050−1055. (19) Ferrin, P.; Mavrikakis, M. J. Am. Chem. Soc. 2009, 131, 14381− 14389. (20) Lin, W.; Stocker, K. M.; Schatz, G. C. J. Phys. Chem. C 2016, 120, 23061−23068. (21) Freund, H. J.; Roberts, M. W. Surf. Sci. Rep. 1996, 25, 225−273. (22) Robota, H. J.; Vielhaber, W.; Lin, M. C.; Segner, J.; Ertl, G. Surf. Sci. 1985, 155, 101−120. (23) Kresse, G.; Furthmuller, J. Phys. Rev. B: Condens. Matter Mater. Phys. 1996, 54, 11169−11186. (24) Kresse, G.; Furthmuller, J. Comput. Mater. Sci. 1996, 6, 15−50. (25) Zhou, X.; Kolb, B.; Luo, X.; Guo, H.; Jiang, B. J. Phys. Chem. C 2017, 121, 5594−5602.

Ni(110) can occur via both associative and redox mechanisms, but only the former is important at low temperature, and (ii) both ER and HA mechanisms occur when H impinges on CO2 adsorbed Ni(110) even at low temperature. We also found that the CO2 coverage on the surface and the H atom velocity affect the overall ratio of ER and HA mechanisms. Furthermore, we have seen that HOCO*, HCOO*, and HCO3* are formed. Overall, the BOMD simulations provide dynamical information that allows us to monitor details of the reaction mechanisms, confirming and extending current understanding of H-assisted CO2 reduction on Ni(110).



ASSOCIATED CONTENT

S Supporting Information *

The Supporting Information is available free of charge on the ACS Publications website at DOI: 10.1021/jacs.7b01538. Computational methods; reaction mechanisms described in the paper; representative trajectories of reflected H, surface H, subsurface H, bulk H, and CO2 desorption; bond distances versus time for representative trajectories (PDF) Movie showing ER (MPG) Movie showing HA1 (MPG) Movie showing HA2 (MPG)



AUTHOR INFORMATION

Corresponding Author

*[email protected] ORCID

Wei Lin: 0000-0002-5046-4765 Present Address †

Department of Chemistry, Suffolk University, Boston, MA.

Notes

The authors declare no competing financial interest.



ACKNOWLEDGMENTS This work is supported by the Air Force Office of Scientific Research through Basic Research Initiative award no. FA955014-1-0053.



REFERENCES

(1) Sanz-Perez, E. S.; Murdock, C. R.; Didas, S. A.; Jones, C. W. Chem. Rev. 2016, 116, 11840−11876. (2) White, J. L.; Baruch, M. F.; Pander, J. E.; Hu, Y.; Fortmeyer, I. C.; Park, J. E.; Zhang, T.; Liao, K.; Gu, J.; Yan, Y.; et al. Chem. Rev. 2015, 115, 12888−12935. (3) Liu, M.; Pang, Y. J.; Zhang, B.; De Luna, P.; Voznyy, O.; Xu, J. X.; Zheng, X. L.; Dinh, C. T.; Fan, F. J.; Cao, C. H. Nature 2016, 537, 382−386. (4) Studt, F.; Sharafutdinov, I.; Abild-Pedersen, F.; Elkjaer, C. F.; Hummelshoj, J. S.; Dahl, S.; Chorkendorff, I.; Norskov, J. K. Nat. Chem. 2014, 6, 320−324. (5) Heine, C.; Lechner, B. A. J.; Bluhm, H.; Salmeron, M. J. Am. Chem. Soc. 2016, 138, 13246−13252. (6) Roiaz, M.; Monachino, E.; Dri, C.; Greiner, M.; Knop-Gericke, A.; Schlogl, R.; Comelli, G.; Vesselli, E. J. Am. Chem. Soc. 2016, 138, 4146−4154. (7) Rettner, C. T.; Auerbach, D. J. Science 1994, 263, 365−367. (8) Quattrucci, J. G.; Jackson, B. J. Chem. Phys. 2005, 122, 074705. (9) Shalashilin, D. V.; Jackson, B.; Persson, M. J. Chem. Phys. 1999, 110, 11038−11046. (10) Dri, C.; Peronio, A.; Vesselli, E.; Africh, C.; Rizzi, M.; Baldereschi, A.; Peressi, M.; Comelli, G. Phys. Rev. B 2010, 82, 165403. D

DOI: 10.1021/jacs.7b01538 J. Am. Chem. Soc. XXXX, XXX, XXX−XXX