Electron to Adsorbate Energy Transfer in Nanoparticles - American

May 30, 2017 - Here, we use ultrafast nonlinear vibrational ... An avenue of research is open where NP size and support thickness, photon energy, and ...
3 downloads 0 Views 3MB Size
Letter pubs.acs.org/JPCL

Electron to Adsorbate Energy Transfer in Nanoparticles: Adsorption Site, Size, and Support Matter Ahmed Ghalgaoui,*,§ Aimeric Ouvrard, Jijin Wang, Serge Carrez, Wanquan Zheng, and Bernard Bourguignon Institut des Sciences Moléculaires d’Orsay (ISMO), CNRS, Univ. Paris-Sud, Université Paris-Saclay, F-91405 Orsay, France ABSTRACT: Confinement of hot electrons in metal nanoparticles (NPs) is expected to lead to increased reactivity in heterogeneous catalysis. NP size as well as support may influence molecule−NP coupling. Here, we use ultrafast nonlinear vibrational spectroscopy to follow energy transfer from hot electrons generated in Pd NP/MgO/Ag(100) to chemisorbed CO. Photoexcitation and photodesorption occur on an ultrashort time scale and are selective according to adsorption site. When the MgO layer is thick enough, it becomes NP size-dependent. Hot electron confinement within NPs is unfavorable for photodesorption, presumably because its dominant effect is to increase relaxation to phonons. An avenue of research is open where NP size and support thickness, photon energy, and molecular electronic structure will be tuned to obtain either molecular stability or reactivity in response to photon excitation.

E

lead to increased catalytic efficiency and selectivity in heterogeneous catalysis.7−12 This is possible because excited photoelectrons can reach adsorbates without energy loss. These two studies have opened promising new possibilities for controlling photochemistry by tuning NP size and light wavelength. The role of wavelength was treated theoretically for well-documented reactions on single crystal surfaces but not yet for NPs.13 There are many other open questions concerning these complex systems: photodesorption may depend on adsorption site and bonding strength; the relative efficiency of single and multiple excitation mechanisms may depend on the system; and the substrate may matter. Such a molecule/ NP/oxide layer system is also relevant for future hybrid devices where a single molecule interconnected to nanostructures will perform electronic functions. In catalysis, more reactivity and selectivity is desired. By contrast, the stability of the molecules during electrical and/or optical excitation is of critical importance for molecular electronics. In this Letter, we will experimentally prove that ultrafast dynamics of CO on Pd NPs grown on a MgO layer depends strongly on the adsorption site, thickness of the oxide layer, and NP size. Using sum frequency generation (SFG) spectroscopy, which has proven to be a very appropriate tool to study the dynamics of molecule−surface interactions,14−18 we are able to follow in real time the dynamics according to adsorption site and NP size, optimize reactivity or stability, and reveal the effect of the support on CO−NP coupling. These results show that while new avenues to control selectivity of catalytic reactions

xcitation of metals with a visible femtosecond laser pulse results in the creation of a transient high density of hot electrons that inelastically scatter through empty hybridized adsorbate−metal orbitals, thereby depositing enough vibrational energy into adsorbates so as to produce photodesorption through repeated electronic excitations.1,2 Two mechanisms have been described: DIMET (desorption induced by multiple electronic transitions) and friction. In each case, multiple hot electrons are involved. In DIMET, each hot electron attachment to an adsorbate triggers energy redistribution between electrons and surface−adsorbate vibrational modes. In friction, energy is transferred through nonadiabatic coupling. Adsorbates may desorb provided that enough energy can be transferred from hot electrons to appropriate nuclear degrees of freedom. Both mechanisms lead to a nonlinear dependence of desorption yield on laser fluence.2,3 This substrate-mediated photoexcitation has been shown to enable catalytic reactions for which an energy barrier must be overcome.4 A recent study of photodesorption from metal nanoparticles (NPs) has shown that electron confinement can allow energy buildup within the hot electron gas and efficient photodesorption by a single electronic excitation of an adsorbate; while in single crystals hot electrons escape very quickly from the surface, resulting in a moderately large “thermal” energy distribution, in NPs, hot electrons are confined and re-excited in the immediate vicinity of the surface, leading to a larger, “athermal” energy distribution. Adsorbates can then desorb by interaction with a single hot electron.5 In addition, hot electrons may couple to plasmon resonance, which enhances electronic excitation. This opens interesting perspectives on plasmonic photochemistry. Another recent study using continuous-light excitation has shown that for small enough NPs (1 ps. NP lattice heating is calculated to be maximal after 1.5 ps using the two-temperature (electrons, phonons) model (Figure 2d). Therefore, the observed large and fast SFG intensity changes can only be ascribed to electronically mediated photodesorption and/or ultrafast changes of electronic excitation.22 In the case of (NO)2/Ag NPs/Al2O3/NiAl(110), two-pulse cross correlation indicates that photodesorption takes place only during the pump pulse (100 fs), which is an argument in favor of the single-excitation mechanism.5 In the present case, the value of 300 fs is an upper limit for the desorption time because the SFG temporal response extends typically over the CO decoherence time T2 (480 fs for 1 ML Pd).20 Therefore, our SFG measurement of a short duration of photodesorption is not conclusive with respect to the single-photon desorption mechanism. Figure 2a displays the CO dynamics as revealed by the facet band intensity as a function of pump−probe delay for 1, 4, and 10 ML of Pd on a 2 ML thick MgO layer. While more refined analysis, including higher temporal resolution, can be obtained from the full simulation of SFG spectra22 (accounting for broadening, shift, and coherent effects visible in Figure 1a), the variation of raw intensity (i.e., peak height) with delay is sufficient to reveal basic time scales. Transients of Figure 2a show the very fast initial laser excitation already shown in

much larger at low-coordinated sites of Pt13 than that at the well-coordinated sites of Pt(111);6,23 the optical transition shifts from 2.05 to 2.85 eV. The second possible explanation is that the CO adsorption energy is site-dependent. It determines the amount of energy that is required to produce photodesorption. Experimental data are available for supported Pd clusters on Al2O3, showing that the adsorption energy is strongly dependent on NP size only in the range of size below ∼300 atoms,24 well below the typical size investigated in the present work (∼820 atoms). Theoretical data25 are also available as a function of adsorption site, but again for small (∼150 atoms) Pd NPs that exhibit mainly (111) facets (as is the case for Fe3O4 or Al2O3 oxide supports). Unfortunately, the support effect was not included in the calculations, while it plays an important role for Pd NPs/MgO/ Ag.20 Therefore, available data concerning small clusters with (111) facets can only be indicative of the present case. For CO on the bridge site located at NP edges between two (111) facets, the adsorption energy was computed to be very high (1.95 eV), as compared to 1.47 eV for bridge sites and 1.15 eV for on-top sites on the (111) facets.25 Accordingly, for CO adsorbed at clusters edges, the Pd−C distance was found to be shorter and the C−O distance slightly longer than that on the regular facets.25 Such a huge difference of adsorption energy would explain alone why CO at edges does not desorb, even if there would be an appropriate DOS to excite CO. Howver, the absence of any spectroscopic response of CO (in addition to the absence of desorption) shows that the band structure is 2668

DOI: 10.1021/acs.jpclett.7b00698 J. Phys. Chem. Lett. 2017, 8, 2666−2671

Letter

The Journal of Physical Chemistry Letters

Figure 3. (a) Excitation and desorption versus Pd deposit on 2 and 10 ML of MgO. See the text for the definition of the phenomenological “electronic excitation” and the steady-state “photoesorption”. Note that the pump energy is not the same for the two MgO thicknesses. Confinement into smaller NPs reduces electronic excitation and photodesorption for thick MgO layers. (b) Calculated reflection coefficient of the bilayer formed by the NP layer (treated as an effective medium) and the MgO layer. The angle of reflection is 67°. All optical parameters are those of bulk materials.

laser fluence is not simple. A quantitative analysis would imply to fit all spectra with an appropriate model that couples electron dynamics, desorption dynamics, and SFG emission.22 For 10 ML of MgO, CO photodesorption decreases by ∼40% from 10 ML (coalesced Pd layer) to 1 ML of Pd, to be compared to ∼0% for 2 ML of MgO (Figure 3a). Electronic excitation also changes much more with Pd deposit in the case of 10 ML of MgO (50%) than that with 2 ML (18%). Comparison of absolute values between 10 and 2 ML of MgO in Figure 3a must take into account the difference of pump energy. The effect of hot electrons from the substrate is minimized at a Pd thickness of 10 ML; therefore, the difference between the photodesorption values at 2 and 10 ML reflects mainly the effect of the pump energy, showing that photodesorption depends nonlinearly on laser energy with a power law of ∼2. For NPs (Pd deposits between 1 and 4 ML), the difference between 10 and 2 ML of MgO is due to both pump energy and NP size. Note that it is necessary to double the pump energy to keep CO excitation at the same level when the volume of the NPs is divided by 4. In summary a significant NP size effect is present, provided that hot electrons from the substrate do not overflow hot electrons from the NP itself. The present result of smaller photoexcitation for smaller NP size for CO/Pd NPs/MgO is opposite from that found for (NO)2/Ag NPs/Al2O3/NiAl(110).5 Several factors can influence the size effect: First, the energy that enters into the system changes strongly between NPs and the continuous layer. The reflection coefficient for our incidence of 67° should vary from 0.97 (the value for the Ag crystal) in the limit of zero Pd deposition to 0.53 (the value for the Pd crystal) for a thick Pd deposit (“thick” at the scale of the optical absorption length of ∼20 ML) (Figure 3b). Thus, we expect much more photoexcitation in the case of the continuous layer than that in the case of NPs, which can explain the large difference in the relaxation dynamics (Figure 2c). Heating lasts much longer, presumably due to a bottleneck in heat transfer through MgO. This may occur if phonon modes of MgO are not resonant with those of Pd and/or Ag. Its effect is stronger as the heat released in palladium is larger. Concerning the comparison between different NP sizes, our optical model (see the Experimental and Computational Methods section) does not predict a

Figure 1b (99.8% purity). Details of MgO film growth and quality optimization are given elsewhere.28 Pd was deposited onto the MgO film, kept at 473 K with equivalent thickness ranging from 0.25 to 10 ML. AFM noncontact and constant height mode was used to characterize the NP density and shape.20 Randomly distributed NPs with high density (1.6 × 1012 NP/cm2) were obtained. It was 10 times larger than that of the MgO(100) single crystal thanks to a higher nucleation point density on the MgO thin layer. A narrow size distribution (around 20%) was observed. A truncated octahedron shape with an aspect ratio of 4:1 (square size/height) was most frequent. The NP square size varied from 2.9 nm for a 0.5 ML equivalent thickness of Pd to 5.8 nm for 4 ML. NPs were coalesced at 10 ML. Optical Methods. An amplified Ti:sapphire laser at 800 nm and 1 kHz is used to pump an optical parametric amplifier (OPA) .The OPA produces tunable IR between 2.5 and 8 μm. The residual beam at 800 nm is split into a pump beam (150 fs, 60 μJ) and a probe beam, the line width and wavelength of which are adjusted by means of a pulse shaper. Vibrationally resolved SFG spectra are obtained by overlapping this ∼2 μJ, 6 cm−1 fwhm “visible” beam and the ∼3 μJ, 150 fs, 150 cm−1 fwhm IR beam onto the sample in a copropagating and collinear configuration. The resulting SFG pulse is spectrally filtered and dispersed by a monochromator on a 512 pixel charge-coupled device camera, which allows us to record SFG spectra of CO in a region of ∼18 nm (∼300 cm−1), centered at ∼700 nm. The spectral resolution in SFG depends on only the bandwidth of the visible laser, and spectra can be acquired in the spectral range covered by the IR pulse without scanning the frequency. All spectra are normalized by reference spectra (ZnSe), allowing direct comparison of band intensities over a 300 cm−1 spectral range. The pump beam is directed onto the sample using the same lens as the probe beams. In this way, it crosses the probe beams on the sample at a very small angle. This ensures that the pump−probe time resolution is not geometrically spoiled and remains close to the pulse duration of the lasers. SFG from the pump and IR beams is emitted at a small angle away from the probe SFG. We use it to achieve spatial and temporal overlap of the pump and IR beams. Spectra are accumulated over typically 105 laser pulses (100 s). 2670

DOI: 10.1021/acs.jpclett.7b00698 J. Phys. Chem. Lett. 2017, 8, 2666−2671

Letter

The Journal of Physical Chemistry Letters

(12) Wodtke, A. M.; Matsiev, D.; Auerbach, D. J. Energy Transfer and Chemical Dynamics at Solid Surfaces: the Special Role of Charge Transfer. Prog. Surf. Sci. 2008, 83, 167−214. (13) Avanesian, T.; Christopher, P. Adsorbate Specificity in Hot Electron Driven Photochemistry on Catalytic Metal Surfaces. J. Phys. Chem. C 2014, 118, 28017−28031. (14) Arnolds, H.; Bonn, M. Ultrafast Surface Vibrational Dynamics. Surf. Sci. Rep. 2010, 65, 45−66. (15) Fournier, F.; Zheng, W.; Carrez, S.; Dubost, H.; Bourguignon, B. Ultrafast Laser Excitation of CO/Pt(111) Probed by Sum Frequency Generation: Coverage Dependent Desorption Efficiency. Phys. Rev. Lett. 2004, 92, 216102. (16) Bonn, M.; Hess, C.; Funk, S.; Miners, J. H.; Persson, B. N. J.; Wolf, M.; Ertl, G. Femtosecond Surface Vibrational Spectroscopy of CO Adsorbed on Ru(001) During Desorption. Phys. Rev. Lett. 2000, 84, 4653. (17) Lane, I. M.; King, D. A.; Liu, Z.-P.; Arnolds, H. Real-Time Observation of Nonadiabatic Surface Dynamics: The First Picosecond in the Dissociation of NO on Iridium. Phys. Rev. Lett. 2006, 97, 186105. (18) Backus, E. H.; Eichler, A.; Kleyn, A. W.; Bonn, M. Real-Time Observation of Molecular Motion on a Surface. Science 2005, 310, 1790−1793. (19) Risse, T.; Shaikhutdinov, S.; Nilius, N.; Sterrer, M.; Freund, H.-J. Gold Supported on Thin Oxide Films: From Single Atoms to Nanoparticles. Acc. Chem. Res. 2008, 41, 949−956. (20) Ouvrard, A.; Ghalgaoui, A.; Michel, C.; Barth, C.; Henry, C. R.; Wang, J.; Carrez, S.; Zheng, W.; Bourguignon, B. CO Chemisorption on Ultrathin MgO-Supported Palladium Nanoparticles. J. Phys. Chem. C 2017, 121, 5551−5564. (21) Ouvrard, A.; Wang, J.; Ghalgaoui, A.; Nave, S.; Carrez, S.; Zheng, W.; Dubost, H.; Bourguignon, B. CO Adsorption on Pd(100) Revisited by Sum Frequency Generation: Evidence for Two Adsorption Sites in the Compression Stage. J. Phys. Chem. C 2014, 118, 19688−19700. (22) Fournier, F.; Zheng, W.; Carrez, S.; Dubost, H.; Bourguignon, B. Vibrational Dynamics of Adsorbed Molecules Under Conditions of Photodesorption: Pump-Probe SFG Spectra of CO/Pt(111). J. Chem. Phys. 2004, 121, 4839−4847. (23) Watanabe, K.; Inoue, K.-I.; Nakai, I. F.; Matsumoto, Y. Nonadiabatic Coupling Between C−O Stretching and Pt Substrate Electrons Enhanced by Frustrated Mode Excitations. Phys. Rev. B: Condens. Matter Mater. Phys. 2010, 81, 241408. (24) Sitja, G.; Le Moal, S.; Marsault, M.; Hamm, G.; Leroy, F.; Henry, C. R. Transition From Molecules to Solid State: Reactivity of Supported Metal Clusters. Nano Lett. 2013, 13, 1977−1982. (25) Yudanov, I. V.; Sahnoun, R.; Neyman, K. M.; Rösch, N.; Hoffmann, J.; Schauermann, S.; Johánek, V.; Unterhalt, H.; Rupprechter, G.; Libuda, J.; et al. CO Adsorption on Pd Nanoparticles: Density Functional and Vibrational Spectroscopy Studies. J. Phys. Chem. B 2003, 107, 255−264. (26) Frischkorn, C.; Wolf, M. Femtochemistry at Metal Surface: Nonadiabatic Reaction Dynamics. Chem. Rev. 2006, 106, 4207−4233. (27) Arbouet, A.; Voisin, C.; Christofilos, D.; Langot, P.; Del Fatti, N.; Vallée, F.; Lermé, J.; Celep, G.; Cottancin, E.; Gaudry, M.; et al. Electron-Phonon Scattering in Metal Clusters. Phys. Rev. Lett. 2003, 90, 177401. (28) Ouvrard, A.; Niebauer, J.; Ghalgaoui, A.; Barth, C.; Henry, C. R.; Bourguignon, B. Characterization of Thin MgO Films on Ag(100) by Low-Energy Electron Diffraction and Scanning Tunneling Microscopy. J. Phys. Chem. C 2011, 115, 8034−8041. (29) Baldelli, S.; Eppler, A. S.; Anderson, E.; Shen, Y. R.; Somorjai, G. A. Surface Enhanced Sum Frequency Generation of Carbon Monoxide Adsorbed on Platinum Nanoparticles Arrays. J. Chem. Phys. 2000, 113, 5432−5438.

Ef fective Model. We model the optical response as a bilayer deposited on a substrate. We approximate the NP layer to a homogeneous layer with effective index29 n = ((1 + 2rq)/(1 − rq))0.5, where r = ((nPd)2 − 1)/((nPd)2 + 1) and q is the fraction of the surface covered by Pd NPs. This model was previously used by the Samorjai group to describe the optical response of platinum particles with a size of 100 nm.29The coverage q of Pd NPs is calculated from a model deduced by AFM experiments.20 Coalescence is obtained for an equivalent thickness of 9 ML of Pd.



AUTHOR INFORMATION

Corresponding Author

*E-mail: [email protected]. ORCID

Ahmed Ghalgaoui: 0000-0002-6978-9690 Aimeric Ouvrard: 0000-0003-3652-1222 Present Address

§ A.G.: Max-Born-Institut für Nichtlineare Optik and Kurzzeitspektroskopie, Max-Born-Strasse 2 a, 12489 Berlin, Germany.

Notes

The authors declare no competing financial interest.



ACKNOWLEDGMENTS We gratefully acknowledge the French National Agency for Research (Agence Nationale de Recherche) for financial support through the PANO program (Project CANA ANR06-NANO-0031).



REFERENCES

(1) Dai, H. L.; Ho, W. Laser Spectroscopy and Photochemistry on Metal Surface; World Scientific: Singapore, 1995. (2) Zhou, X.-X.; Zhu, X.-Y.; White, J. M. Photochemistry at Adsorbate/Metal Interfaces. Surf. Sci. Rep. 1991, 13, 73−220. (3) Menzel, D. Electronically Induced Surface Reactions: Evolution, Concepts, and Perspectives. J. Chem. Phys. 2012, 137, 091702. (4) Bonn, M.; Funk, S.; Hess, C.; Denzler, D. N.; Stampfl, C.; Scheffler, M.; Wolf, M.; Ertl, G. Phonon-Versus Electron-Mediated Desorption and Oxidation of CO on Ru(0001). Science 1999, 285, 1042−1045. (5) Kim, K. H.; Watanabe, K.; Mulugeta, D.; Freund, H.-J.; Menzel, D. Enhanced Photoinduced Desorption from Metal Nanoparticles by Photoexcitation of Confined Hot Electrons Using Femtosecond Laser Pulses. Phys. Rev. Lett. 2011, 107, 047401. (6) Kale, M. J.; Avanesian, T.; Xin, H.; Yan, J.; Christopher, P. Controlling Catalytic Selectivity on Metal Nanoparticles by Direct Photoexcitation of Adsorbate-Metal Bonds. Nano Lett. 2014, 14, 5405−5412. (7) Wolf, M.; Ertl, G. Surface Science: Electron Dynamics at Surface. Science 2000, 288, 1352−1353. (8) Studt, F.; Abild-Pedersen, F.; Bligaard, T.; Sørensen, R. Z.; Christensen, C. H.; Nørskov, J. K. Identification of Non-Precious Metal Alloy Catalysts for Selective Hydrogenation of Acetylene. Science 2008, 320, 1320−1322. (9) Marimuthu, A.; Zhang, J.; Linic, S. Tuning Selectivity in Propylene Epoxidation by Plasmon Mediated Photo-Switching of Cu Oxidation State. Science 2013, 339, 1590−1593. (10) Wodtke, A. M.; Tully, J. C.; Auerbach, D. J. Electronically NonAdiabatic Interactions of Molecules at Metal Surfaces: Can we Trust the Born-Oppenheimer Approximation for Surface Chemistry? Int. Rev. Phys. Chem. 2004, 23, 513−539. (11) Tully, J. C. Chemical Dynamics at Metal Surfaces. Annu. Rev. Phys. Chem. 2000, 51, 153−178. 2671

DOI: 10.1021/acs.jpclett.7b00698 J. Phys. Chem. Lett. 2017, 8, 2666−2671