Ultrafast Vibrational Energy Transfer in the Layers of D2O and CO on

Apr 15, 2009 - Graduate School of Science, Department of Chemistry, Kyoto UniVersity, Kyoto 606-8502, Japan, and. PRESTO-JST, 4-1-8 Honcho Kawaguchi ...
0 downloads 0 Views 1MB Size
Download PDF
11712

J. Phys. Chem. C 2009, 113, 11712–11719

Ultrafast Vibrational Energy Transfer in the Layers of D2O and CO on Pt(111) Studied with Time-Resolved Sum-Frequency-Generation Spectroscopy† Masashi Nagao,‡,§ Kazuya Watanabe,*,⊥,¶ and Yoshiyasu Matsumoto*,⊥,‡ National Institutes of Natural Sciences, Institute for Molecular Science, Okazaki, Aichi 444-8585, Japan, Graduate School of Science, Department of Chemistry, Kyoto UniVersity, Kyoto 606-8502, Japan, and PRESTO-JST, 4-1-8 Honcho Kawaguchi, Saitama 332-0012, Japan ReceiVed: February 26, 2009; ReVised Manuscript ReceiVed: April 1, 2009

Ultrafast dynamics of vibrational energy transfer in overlayers of D2O and CO on Pt(111) have been investigated by femtosecond time-resolved (TR) IR-visible sum-frequency-generation (SFG) spectroscopy under ultrahighvacuum conditions. About 10 layers of D2O ice were epitaxially grown on c(4 × 2)-CO/Pt(111). The surface was excited by subpicosecond laser pulses, and subsequent energy transfer through low-frequency modes of adsorbates was monitored in terms of peak shifts and broadenings of C-O and O-D stretching bands in SFG spectra as a function of the pump-probe delay. Because D2O ice forms islands, there are two types of CO: one interacting with D2O and the other free from D2O. Simulations of the TR-SFG spectra by using a phenomenological model for the energy-transfer dynamics indicate that the coupling rate of perturbed CO is larger than that of free CO by a factor of 1.7; this is probably because CO 2π* states shift toward the Fermi level due to interaction with D2O. Two isolated bands at 2668 and 2713 cm-1 were assignable to the OD stretching bands of D2O directly interacting with CO at the D2O/CO interface and D2O at the vacuum/ice interface, respectively. Analysis of the temporal spectral changes of free D2O by using a diffusive thermal transport model indicates that heat transfer through low-frequency phonons of the ice layers occurs within 3 ps; this is substantially faster than the pulsed laser-induced melting of thin ice films reported previously. 1. Introduction Vibrational excitation and relaxation at a surface are of great interest in surface chemistry and heterogeneous catalysis. Carbon monoxide (CO) is a prototype adsorbate, and its vibrational dynamics on metal surfaces have been most extensively studied among various adsorption systems.1-7 Early studies on the lineshape analysis of CO stretching bands revealed that the creation of electron-hole pairs plays an important role in energy dissipation of the internal stretching mode8 and dephasing is induced by anharmonic couplings of the CO stretching mode to low-frequency modes: frustrated rotation and translation of CO.9 Time-resolved (TR) infrared (IR) reflection and IR-visible sum-frequency-generation (SFG) spectroscopy with picosecond laser pulses have been successfully applied to the population dynamics of the CO internal stretching mode as well as the coupling of substrate electrons and phonons to the CO frustrated modes.10-15 More recently, ultrafast heating of the frustrated modes on a subpicosecond time scale induced by femtosecond near-IR pulses has been elucidated by using femtosecond TR-SFG.16-22 Hot electrons with transient temperatures higher than a few thousand Kelvin efficiently couple with the frustrated modes, ultimately leading to nuclear motions: hopping and desorption of adsorbates.23-26 In spite of the abundant pieces of information on the structure and dynamics of CO on metal surfaces, the effects of coadsor†

Part of the “Hiroshi Masuhara Festschrift”. * E-mail: [email protected] (K.W.), matsumoto@ kuchem.kyoto-u.ac.jp (Y.M.). ‡ Institute for Molecular Science. § Present address: Nanomaterials Technology Development Center, Canon Inc., 3-30-2 Shimomaruko, Ohta-ku, Tokyo 146-8501, Japan. ⊥ Kyoto University. ¶ PRESTO-JST.

bates on the ultrafast vibrational dynamics have been less explored. Because interadsorbate interaction plays a crucial role in surface reactions of practical importance, it is important to explore how coadsorbates influence the vibrational dynamics of CO interacting with the substrate in highly nonequilibrium conditions. In this study, we applied femtosecond TR-SFG to CO on Pt(111) coadsorbed with D2O. The coadsorption of CO and water on Pt(111) in an ultrahigh vacuum (UHV) has been investigated in comparison with a Pt(111)-aqueous electrochemical interface.27-29 The tactic of “UHV electrochemical modeling” has successfully proven structural comparability between the two systems. Moreover, the greater interest has been centered on solid ice in various fields of science because of its importance from fundamental, environmental, and industrial points of view.30 IR reflection absorption spectroscopy (IRAS) characterized how the CO stretching band is affected by water coadsorbates.27,31 At low and intermediate coverages of CO on Pt(111), i.e., θCO e 0.25 ML (1 ML) 1.5 × 1015 molecules cm-2), the band intensity of the atop CO decreases, while that of the bridge CO increases as water coverage increases, indicating that coadsorption of water shifts the adsorption site of CO. In contrast, at higher coverages of CO, i.e., θCO g 0.5 ML, the atop CO band at 2105 cm-1 progressively decreases in intensity, while a new broader band appears at 2077 cm-1 and grows with water coverage. This “titration-like” response to the water dosage was interpreted as follows: water molecules cannot bind directly to Pt atoms, but they adsorb on the CO layer to form islands. The broadening and red shift of the atop CO band are manifestations of the interaction between CO and water. A comparison of the vibrational spectra of ice on Pt(111) with those of ice on CO/ Pt(111) indicates that ice overlayers grow epitaxially on the monolayer of CO. Summarizing these findings, we depict a

10.1021/jp901793q CCC: $40.75  2009 American Chemical Society Published on Web 04/15/2009

Ultrafast Vibrational Energy Transfer in D2O and CO

J. Phys. Chem. C, Vol. 113, No. 27, 2009 11713 2. Experimental Procedure

Figure 1. Structural model of D2O adsorbed on CO/Pt(111). There are two kinds of CO: CO perturbed by interaction with D2O, denoted as p-CO, and CO free from interaction with D2O, denoted as f-CO. Similarly, OD bonds directly interacting with CO at the CO/D2O interface are denoted as p-OD, while those at the vacuum/ice interface are denoted as f-OD.

schematic picture of the coadsorption system, D2O/CO/Pt(111), in Figure 1. Most of CO molecules interact with D2O molecules, but some CO molecules are free from D2O because D2O crystals grow as islands on this surface. From now on, CO perturbed by interaction with water is denoted as p-CO, whereas CO free from interaction is denoted as f-CO. Electronic excitation by femtosecond laser pulses induces various ultrafast nuclear dynamics of adsorbates: vibrational excitation, desorption, and chemical reactions. These processes have been explained in terms of electronic friction or the multiple electronic transition model.32 In both of the mechanisms, the electronic structure of adsorbates is a key factor for understanding the ultrafast dynamics. Because the frequency of the CO internal stretching mode is a measure of the occupation in the 2π* band,33 a marked red shift of the p-CO internal stretching mode indicates that a larger fraction of electron is back-donated into the 2π* band, weakening the CO bond. This is probably due to the fact that the CO 2π* band is shifted toward the Fermi level or broadened or both by interaction with D2O. Hence, it is interesting to know how the electronic structural change due to interactions with coadsorbates influences the dynamical response to electronic excitation by ultrashort laser pulses. Few studies have been performed on vibrational relaxation and energy transfer in water/CO overlayers. Kubota et al. investigated heat flows in the overlayers by using picosecond TR-SFG.34,35 They prepared several tens of ice layers epitaxially grown on a monolayer of CO on Pt(111). The Pt surface was transiently heated by 35-ps pulses in near-IR, and temporal variations in CO and OD stretching bands were recorded. While the response of the CO stretching band is almost instantaneous to the pump pulse, a significantly slower transient response ranging from 200 to 400 ps was observed in the OD stretching band. Their analysis with a macroscopic thermal diffusion model revealed that the observed transient response in the OD stretching band is significantly slower than that expected from diffusive thermal transport in ice. To obtain a deeper insight into the vibrational energy-transfer dynamics at the D2O/CO interface, it is crucial to study with a better time resolution. In this paper, we report on subpicosecond transient responses of the CO and OD stretching bands of the coadsorption system, D2O/CO/Pt(111), which are induced by a 150-fs pump laser pulse at 400 nm and probed by TR-SFG. We show that p-CO couples with substrate hot electrons more efficiently than f-CO. In addition, we show that vibrational energy transfer in the ice overlayers manifests itself in transient spectral changes in the OD stretching band of D2O at the vacuum/ice interface.

The experiments were performed in an UHV chamber at a base pressure of less than 2 × 10-10 Torr. The chamber is equipped with a cylindrical electron energy analyzer with an electron gun for Auger electron spectroscopy (AES), low-energy electron diffraction (LEED) optics, and a quadrupole mass spectrometer. A Pt(111) surface with dimensions of 10 mm (diameter) × 1.0 mm (thickness) was cleaned by repeating a cycle of Ar ion sputtering, annealing, oxidation, and flashing.36 Impurities (C, O, and S) at the surface were confirmed to be less than 3% with AES. The surface temperature was measured with a chromel-alumel (K-type) thermocouple spot-welded to the side of the crystal. The sample could be cooled to 83 K with liquid N2. Deuterated water molecules (D2O, 99.8% purity, Acros Organics) was degassed by the freeze-pump-thaw method prior to introduction to the UHV chamber. Ice layers were epitaxially grown on CO/Pt(111) in the following procedure. First, dosing the clean Pt surface with CO at 83-90 K was followed by annealing at 260 K for preparation of a CO monolayer with a superstructure of c(4 × 2)-CO that was confirmed with LEED. Then, D2O was further dosed onto the CO-covered surface at 140 K to form crystalline ice layers.27,31,37,38 The coverage of D2O was determined by a comparison of the integrated area of the thermal desorption peak of D2O with that of one bilayer of ice. Laser pulses (λ ) 800 nm) with a typical duration of 130 fs from a Ti:sapphire regenerative amplifier (Spectra Physics, Spitfire-pro, 1 kHz repetition rate) were divided into two beams. One of the beams was fed into an optical parametric generation/ amplifier (Light Conversion, TOPAS) that produced signal and idler beams. Difference frequency mixing of the signal and idler beams by using a AgGaS2 crystal generated broad-band IR pulses with a spectral width of ∼150 cm-1, a pulse energy of 2-5 µJ, and a pulse duration of ∼150 fs. The wavelength of IR pulses was tunable in the range of 1-10 µm. The other output of the regenerative amplifier was frequency-doubled by a BBO (β-BaB2O4) crystal to provide a pump pulse (λ ) 400 nm; duration ) 150 fs) for TR measurements. Residual fundamental pulses (λ ) 800 nm) penetrating through the BBO crystal were stretched to a pulse duration of ∼3 ps with the aid of a pulse stretcher consisting of a grating, a cylindrical lens, and a slit. The stretched fundamental and broad-band IR pulses were used for SFG probe pulses.39 The IR and “visible” (λ ) 800 nm) probe beams both in p polarization were focused and overlapped spatially as well as temporally onto the sample surface. While the angle of incidence of the visible beam was 68° from the surface normal, the IR beam was incident with an angle of incidence larger than that of the visible beam by ∼3°. The absorbed fluence of the pump pulse was 8 J m-2. SFG signals passing through an iris and filters for the reduction of scattering light of pump pulses were dispersed in a spectrometer and detected with a CCD camera (Princeton) as a function of the peak-to-peak time delay between the pump (λ ) 400 nm) and the IR probe pulses, td. 3. Results 3.1. Coadsorption Effects on SFG Spectra. At θCO ) 0.5 ML, CO molecules on Pt(111) adsorb both on the atop and bridge sites to form a c(4 × 2) superstructure. In previous reports,40-42 SFG spectra of the CO stretching bands of atop and bridge CO molecules were recorded at ∼2100 and ∼1850 cm-1, respectively. The SFG signal intensity from bridge CO was much weaker than that from atop CO; this makes it difficult

11714

J. Phys. Chem. C, Vol. 113, No. 27, 2009

Nagao et al.

Figure 2. SFG spectra (red line) of D2O/CO/Pt(111). The coverages of CO and D2O are 0.5 and 5 ML, respectively. The intensity distributions of IR pulses are depicted by blue lines. The SFG signals are normalized by the intensities of corresponding broad-band IR spectra.

to record subtle transient changes in the SFG spectra of bridge CO with high quality. Thus, we focus only on the atop CO band in this paper. Figure 2 shows the SFG spectra measured at 2000-3000 cm-1 for the Pt(111) surface covered with CO and D2O. The coverage of CO was 0.5 ML, and the coverage of D2O, θD, was 5 ML. When D2O was adsorbed on the 0.5 ML CO precovered Pt(111) surface, the CO stretching band at 2095 cm-1 broadened and another band appeared at 2068 cm-1. When D2O coverage was further increased, the original band at 2095 cm-1 disappeared and only the red-shifted band remained visible. If ice grew homogeneously in a layer-by-layer fashion, the stretching band of atop CO should have remained as a single perturbed peak at θD g 1 ML. In contrast, the atop CO band split into two bands. The splitting could be accounted for only by the island growth of D2O ice layers on a CO monolayer; only CO adsorbates covered with islands of D2O ice are perturbed. This “titration behavior” is similar to changes in IRAS spectra measured previously.27,31,43 Weaver and co-workers27,43 reported that the stretching band of atop CO at 2105 cm-1 observed for 0.55 ML CO/Pt(111) broadened and shifted to 2077 cm-1 as a result of water coadsorption. Therefore, we assigned the red-shifted band at 2068 cm-1 to the stretching band of CO interacting with D2O. Hereafter, we denote the CO stretching band at 2095 cm-1 free from perturbation of D2O as νf-CO and that perturbed by D2O at 2068 cm-1 as νp-CO. When D2O was adsorbed on the 0.5 ML CO/Pt(111) surface, several bands corresponding to OD stretching bands of D2O appeared at 2200-2800 cm-1. Two broad bands at 2278 and 2472 cm-1 are assignable to the OD symmetric and asymmetric stretching modes of crystalline ice, respectively.34,44,45 In addition to these strong bands, two weak bands appeared at 2675 and 2720 cm-1. According to IRAS studies on Pt(111) surfaces coadsorbed with CO and D2O,27,31,43 we assigned the bands at 2675 and 2720 cm-1 to the OD stretching mode of D2O directly interacting with CO molecules (νp-OD) at the ice/CO interface and that of non-hydrogen-bonded species (νf-OD) at the vacuum/ ice interface, respectively. 3.2. TR-SFG Spectra. Upon excitation with pump pulses, SFG spectra showed transient variations as a function of the pump-probe delay, td. In this work, we focus on the transient responses in two frequency regions: CO stretching (2000-2125 cm-1) and isolated OD stretching (2600-2750 cm-1) because these bands are sensitive to energy-transfer dynamics. Figure 3 shows a series of TR-SFG spectra (red lines) of νf-CO and νp-CO. It is evident that both of the CO bands show time-dependent

Figure 3. Experimental (red) and simulated (black) TR-SFG spectra of CO stretching bands for D2O/CO/Pt(111). The coverages of CO and D2O are 0.5 and 10 ML, respectively. An absorbed fluence of the pump pulse (λ ) 400 nm) is 8 J m-2 per pulse. Pump-probe delay times, td, are indicated in the figure. See the text for details of the simulation procedures by using parameters listed in Table 1.

Figure 4. Temporal variations of the peak frequencies of (a) νf-CO and (b) νp-CO in the TR-SFG spectra. Experimental results are plotted by open circles, and calculated results with various electronic coupling constants, τ, are drawn by colored lines.

variations in frequency, width, and intensity. The transient peak shifts of the CO stretching bands estimated from the experimental data are plotted in Figure 4. Both of the bands red shift near td ) 0 ps and shift back to the initial frequencies at longer delay times. The peak of νp-CO shows a larger frequency shift at td ∼ 0 ps and faster recovery than that of νf-CO. Pump-pulse irradiation induced more subtle changes in the spectral features of OD stretching bands. Figure 5 shows SFG spectra (red lines) of νf-OD and νp-OD as a function of td. Both of the peaks decrease in intensity and show small peak shifts (∼0.5 cm-1) in 0 < td < 3 ps. Figure 6 shows the transient intensity variations in detail. The intensities of both of the peaks decrease to ∼75% of those without pump pulses at td ∼ 0 ps and recover slowly at larger td. Because the OD bands, particularly νp-OD, are broad and overlap each other, it was necessary to fit all of

Ultrafast Vibrational Energy Transfer in D2O and CO

J. Phys. Chem. C, Vol. 113, No. 27, 2009 11715

TABLE 1: Parameters Used for Simulations of νCO SFG Spectra νf-CO νp-CO

ω(T0) (cm-1)

Γ(T0) (THz)

R (cm-1/K)

β (THz K-1)

τj (ps)

|φp-CO - φf-CO| (rad)

r

2095 2072

0.83 1.60

-0.018 -0.014

0.007 0.007

1.7 1.0

0.29 0.29

1.8 1.8

the spectra with an appropriate model for quantitative analysis, which is described in the next section. 4. Analysis and Discussion 4.1. Energy-Transfer Model. Before we go into the details of the quantitative analysis of TR-SFG spectra, here we outline a phenomenological model for energy-transfer dynamics at D2O/ CO/Pt(111), as schematically depicted in Figure 7. Pump pulses

Figure 5. Experimental (red) and calculated (black) TR-SFG spectra in the OD stretching region. The adsorbate coverages and excitation conditions are the same as those in Figure 3. The pump-probe delay times, td, are indicated in the figure.

Figure 6. Temporal variations of the peak intensities of (a) νf-OD at 2708 ( 4 cm-1 and (b) νp-CO at 2674 ( 4 cm-1 in the TR-SFG spectra. Experimental results are plotted by open circles. Simulation results with eq 14 are also depicted for various k′: k′ ) 1000 (blue), 740 (green), 420 W K m-1 (red line). Three simulation curves for p-OD coincide with each other because variation in the thermal conductivity does not affect Tp-OD.

excite electrons in the metal. Deposited energy is transferred to lattice phonons as well as nuclear motions of adsorbates. It has been well established that hot electrons excite frustrated translation and rotation of adsorbed CO.12,16,20-22 Red shifts and broadenings of νCO are manifestations of the excitation of frustrated motions because of anharmonic couplings between the CO internal stretching and frustrated modes. The coupling strengths depend on the populations of frustrated modes.1,46,47 In previous studies of laser-induced surface vibrational dynamics,32,48 the population of a frustrated mode is characterized by a mode-dependent effective temperature under the assumption of local thermal equilibrium of the system. Because the anharmonic coupling between the frustrated rotation and CO stretching modes is moderate on Pt(111),21 we consider only the coupling of CO stretching with the frustrated translation and denote the effective temperatures for the frustrated modes of f-CO and p-CO as Tf-CO and Tp-CO, respectively. Similarly, the observed transient changes in νp-OD are likely caused by anharmonic couplings between the OD stretching mode and ice phonon modes at the interface. We represent the population of ice phonon modes at the D2O/CO interface by an effective temperature for p-OD, Tp-OD. Because the frustrated mode of p-CO also couples with the ice phonon modes at the interface, we assume that a local equilibrium between p-CO and p-OD is rapidly established: Tp-CO is equal to Tp-OD. Further propagation of the excess energy toward the vacuum side is dominated by phonon-phonon couplings in the ice layer. The pump-induced variation of νf-OD is a manifestation of the heat transport up to the ice/vacuum interface, and its spectral change is due to anharmonic couplings of νf-OD with ice phonon modes, whose population is characterized by T f-OD. Using the phenomenological model described here, we simulated the observed TR-SFG spectra, as described in detail in the following sections. 4.2. Fitting of SFG Spectra. For quantitative analysis of the transient SFG spectra, we adopted the method utilized in earlier studies,10,16-18,20-22,49 i.e., Fourier transformation of time-

Figure 7. Schematic model for energy-transfer dynamics at D2O/CO/ Pt(111). A very high transient electronic temperature, Te, is achieved by the laser pulse excitation of substrate electrons. Hot electrons couple with adsorbate CO, giving rise to the local adsorbate temperatures Tp-CO and T f-CO. The electron coupling strength is characterized by the electron coupling rate, 1/τj (j ) p or f). The energy deposited on p-CO further propagates into the ice film, leading to the phonon excitation at the vacuum/ice interface, which is characterized by T f-OD.

11716

J. Phys. Chem. C, Vol. 113, No. 27, 2009

Nagao et al.

dependent polarization. We assumed that each spectrum in Figures 3 and 5 is composed of two vibrational modes and the nonlinear polarization of the system is given by a linear combination of their polarizations. Then, the second-order nonlinear polarization induced by IR and visible pulses is given by

P(2)(t) ∝

∑ AiPiIR(t) E(t) cos(ωvist + φi)

(1)

i

Here, PIR i (t) stands for an IR-induced vibrational polarization of an ith vibrational mode; the visible field is assumed to be quasi-monochromatic with an angular frequency of ωvis and a temporal envelope of E(t); Ai and φi are the relative amplitude and phase of each nonlinear polarization, respectively. We neglected a nonlinear polarization term due to vibrationally nonresonant transitions of the metal substrate because its amplitude is negligibly small on Pt(111) at the current experimental conditions. The temporal profile of PIR i (t) is described in terms of the density matrix of a two-level system: a vibrational ground state |0〉 and an excited state |1〉

PiIR(t) ∝ Tr[Fˆ i(t) µˆ ]

(2)

where Fˆ i(t) is the density operator for the ith mode and µˆ is the dipole operator. We calculate PIR i (t) by numerically solving the following equations of motion with the fourth-order Runge-Kutta algorithm

∂F10 i ) [-iωi(Ti) - Γi(Ti)]F10 + V10(F11 - F00) (3) ∂t p ∂F11 1 (4) ) (V10F01 - F10V01) - γiF11 ∂t ip F11 + F00 ) 1 (5) F10 ) F*01 V10 )

V*01

)

(6)

IR -µ(i) 10E (t)

(7)

where γi is a population relaxation rate and F10 and matrix elements of Fˆ i and µˆ . A resonant frequency, ωi, and a dephasing rate, Γi, are assumed to depend on the temperature of the frustrated mode, Ti. Then SFG spectra are obtained by Fourier transformation of P(2)(t) µ(i) 10 are

ISFG(ω) ∝

|∫

dt e-iωtP(2)(t)

|

2

(8)

(i) We determined the parameters ωi(Ti), Γi(Ti), µ10 Ai, and φi by fitting the TR-SFG spectra experimentally observed with simulations as described in the following sections. 4.3. Energy Transfer at a Metal/CO Interface. We employed a two-temperature model for simulating the transient changes of the substrate electron temperature, Te, and the lattice temperature, Tl, upon laser heating.12,32,50,51 The time profiles of Te and Tl were estimated by solving the following coupled equations numerically:

Ce

∂Te ) κ ∇2Te - g(Te - Tl) + S(z, t) ∂t ∂Tl Cl ) g(Te - Tl) ∂t

(9) (10)

Here, Ce, Cl, κ, and g are the heat capacities of electrons and lattices, the thermal conductivity of electrons, and the electron-

Figure 8. (a) Pump-induced variations of electron (blue) and lattice (red) temperatures estimated with the two-temperature model (eqs 9 and 10). The absorbed fluence is 8 J m-2. (b) Simulation results for the temporal variations of adsorbate temperatures by using eqs 9-11. Temperatures of Tp-CO (red), Tf-CO (green), and Tf-OD (blue) are estimated by using the calculated substrate temperatures depicted in part a and the best-fit parameters listed in Tables 1 and 2.

phonon coupling constant, respectively. The term S(z,t) denotes an energy source provided by laser pulse excitation. All of the parameters necessary for calculating the temperatures of the Pt substrate are available in the literature.12 The calculation result is depicted in Figure 8a. The transient temperature of the CO frustrated mode was calculated by using the heat-transfer equation32,48

dTj-CO Te - Tj-CO ) dt τj

(11)

Here, j ) f or p and 1/τj is the electron coupling rate for j-CO. We neglected the coupling between the CO frustrated mode and lattice phonons because the coupling with hot electrons likely dominates over that with lattice phonons in the dynamics in a subpicosecond range.21,48 Recently, Ueba and Persson showed that anharmonic couplings between the frustrated rotation and translation modes play an important role in ultrafast heating of adsorbates.48,52 However, the couplings among the frustrated modes were neglected here because the employed laser fluence was moderate in comparison with those in the previous studies of hopping or desorption of adsorbate induced by irradiation of high fluence laser pulses.20,21 Furthermore, we neglected energy transfer from the CO monolayer to the ice layer because the total amount of heat flow is less than 10% of the deposited total energy by the pump pulse, as will be shown later. The anharmonic couplings with the frustrated mode lead to variations in the CO stretching frequency, ωi, and the dephasing rate, Γi. The amounts of shift and broadening depend on the population of the frustrated mode characterized by the effective temperature, Ti (i ) f-CO or p-CO); these can be approximated by the linear dependence12,46,53

ωi(Ti) ) ωi(T0) + Ri[Ti(t) - T0]

(12)

Γi(Ti) ) Γi(T0) + βi[Ti(t) - T0]

(13)

where Ri and βi are mode-specific constants and T0 is the idle temperature before pump-pulse irradiation, i.e., 83 K. Fittings of the SFG spectra were carried out as follows. First, (i) by minimizing the we determined ωi(T0), Γi(T0), φi, and Aiµ10 deviation of a simulated SFG spectrum from the experimental data without pump-pulse irradiation. Because a broad wing of hydrogen-bonded OD stretching bands overlaps with νp-CO and νf-CO, we incorporated its effect in the simulation by adding an IR-induced instantaneous polarization to PIR i (t). Table 1 lists the

Ultrafast Vibrational Energy Transfer in D2O and CO

J. Phys. Chem. C, Vol. 113, No. 27, 2009 11717

TABLE 2: Parameters Used for Simulations of νOD SFG Spectra νf-OD νp-OD

ω(T0) (cm-1)

Γ(T0) (THz)

R (cm-1/K)

β (THz/K)

|φp-OD - φf-OD| (rad)

r

k′ (W K m-1)

2713 2668

0.97 3.24

-0.01 0.01

0.007 0.012

0.39 0.39

-3.6 -3.6

740 740

obtained spectroscopic parameters for the CO stretching modes, ωi(T0), Γi(T0), and |φp-CO - φf-CO|, and the relative amplitude of f-CO two modes r ) Ap-COµp-CO 10 /Af-COµ10 . The dephasing rate of p-CO is about twice that of f-CO, indicating that hydrogen-bonding configurations around p-OD are inhomogeneous or dephasing of the stretching mode is enhanced owing to coupling with the interface phonons of the ice overlayer or both. There is a definite phase difference between the two CO stretching modes, implying that detuning to the electronic resonance at ωvis for p-CO is different from that for f-CO. Second, Ri and βi were determined by fitting the TR-SFG spectra at td ) 5, 10, and 30 ps. Because Te, Tl, and T j-CO are equilibrated to each other at these delay times, SFG spectra no longer depend on τj; this enables us to optimize Ri and βi. Third, by using the optimized values of Ri and βi, the SFG spectra at td ) 0-1 ps were simulated to determine τj. These simulation results are shown in Figures 3 and 4, and the obtained parameters are listed in Table 1. The simulated spectra reasonably reproduce the observed ones. However, some discrepancies remain on the intensity changes in the spectra: in particular, the observed spectra at td ∼ 0 ps have larger intensities than the simulations. The discrepancies can be reduced if we include an increase in the Raman tensors induced by changes in the electronic structure of the Pt substrate immediately after laser excitation.21 The CO stretching frequency of atop CO on Pt(111) changes linearly with temperature at near 150 K with a constant of proportionality of -0.024 cm-1/K as a result of anharmonic coupling between the CO stretching and the frustrated translational modes.12,53 The value of Rf-CO ) -0.018 cm-1/K obtained in this study is close to the reported one. This indicates that the assumption made in the simulations, including only the frustrated translation mode for the exchange mode, is reasonable. A slightly smaller absolute value of Rp-CO indicates an increase in the force constant of the frustrated translation mode due to interaction with D2O. Figure 8b shows Tj-CO(t) numerically simulated with the parameters in Table 1. Germer et al.12 reported an electron coupling time of 2 ( 1 ps for CO/Pt(111); this is in good agreement with τf-CO ) 1.7 ps in this study. However, τp-CO ) 1.0 ps is shorter than τf-CO by a factor of ∼2. This indicates that hot electrons couple more strongly with the low-frequency modes of p-CO than with those of f-CO. Kizhakevariam et al.27 reported that water adsorption onto CO/Pt(111) yields a substantial decrease in the work function (∼0.3 eV) caused by changes in the electronic structure of CO. Because the red shift of the CO stretching mode is correlated with occupation of the CO 2π* band,54 the 2π* band of p-CO may become closer to the Fermi level (or more broadened) than that of f-CO as a result of interaction with D2O. Because the frictional coupling strength depends on the degree of charge transfer upon adsorption,32 the enhanced occupation in the 2π* band is likely responsible for the relatively large coupling with hot electrons for p-CO. 4.4. Energy Flow in the Ice Film. Ice possesses translational phonon modes with frequencies lower than ∼400 cm-1.30 These low-frequency ice phonons play a crucial role in the transport of excess vibrational energy from the CO monolayer into the ice film. We assume that the degree of phonon excitation is represented by transient local temperatures in the ice layer, Tw,

which is a function of td and the distance from the CO/D2O interface, z. Although it is widely known that the conventional Fourier law of heat transport breaks down when the medium size is smaller than phonon mean free paths,55 we simulated Tw(z,td) by adopting a one-dimensional thermal diffusion equation as a zeroth-order approximation

Cp(Tw)

[

∂Tw(z, td) ∂Tw(z, td) ∂ ) k(Tw) ∂td ∂z ∂z

]

(14)

where Cp(Tw) is the heat capacity and k(Tw) is the thermal conductivity. We used the molar heat capacity of bulk D2O ice, which varies almost linearly with temperature from 1.7 to 4.2 J K-1 mol-1 from 100 to 250 K,30 while the thermal conductivity is assumed to be inversely proportional to the temperature as

k)

k' T

(15)

which has been proven to be valid for bulk ice at above 60 K with k′ ) 651 W K m-1.56 The D2O/CO interface is positioned at z ) 0, and we assumed that Tw(0,td) ) Tp-OD(td). The thickness of the ice film was assumed to be 3.7 nm (10 ML) and homogeneous, and the temperature at the vacuum/ice interface, Tw(z ) 3.7 nm, td), was assumed to equal Tf-OD. Then the proportionality constant k′ was determined by fitting of the TRSFG spectra. The heat capacity of bulk ice should be used in the simulation with caution because the heat capacity on the nanometer scale may differ from that of bulk. For insulators, the differences are mainly due to the changes in the phonon density of states with reduction of the system size. A recent investigation on the heat capacity of water confined in nanopores revealed that its heat capacity can be larger than that of bulk water by a factor of 1.4.57 Because the heat capacity of ice on the nanometer scale is not known, we used bulk heat capacity as a zeroth-order approximation. We simulated the TR-SFG spectra in the range of OD stretching modes (Figure 5) in almost the same manner as that for CO stretching modes. The frequency and dephasing rates of p-OD and f-OD were determined by simulating the SFG spectra without pump-pulse irradiation. The obtained spectroscopic parameters are listed in Table 2. Note that the sign of the relative amplitude between f-OD and p-OD is minus, indicating that the induced polarization Aiµ(i) 10 of f-OD is opposite to that of p-OD. This is because p-OD points to the CO monolayer, while f-OD points to the vacuum, as depicted in Figure 1. Low-frequency phonons excited at the interfaces may couple with OD stretching vibrations, leading to peak shifts and broadenings of the bands. We adopt the same functional forms in eqs 12 and 13 for the temperature dependence of OD stretching bands. The parameters of Ri and βi for each mode were determined by fitting globally the TR-SFG spectra of νi (i ) p-OD or f-OD) at td ) 5, 10, and 30 ps. The temperature Tp-OD was assumed to be the same as Tp-CO, which was determined by eq 11, while Tf-OD was estimated by solving eq 14 numerically. Then, by using these parameters, we determined k′ by fitting globally TR-SFG spectra for td ) -0.5-5.0 ps. The best-fit results are depicted in Figures 5 and 6, and the fitting parameters are listed in Table 2. As in the case of the

11718

J. Phys. Chem. C, Vol. 113, No. 27, 2009

fittings of CO stretching bands, the transient increase in the intensity observed at td ∼ 0 ps deviates from the simulation. This could be also due to the effect of substrate electronic excitation on the surface Raman tensors. Figure 8b shows the time profile of Tf-OD calculated by eq 14 with the optimized parameters. It is evident that the temperature rise at the vacuum/ ice interface occurs within 3 ps. The best-fit value of k′ ) 740 W K m-1 reasonably agrees with the estimated one for bulk ice, 651 W K m-1. Thus, this may indicate that the energy-transfer dynamics even in the ultrathin ice film with a thickness of 3.7 nm are well described by the one-dimensional thermal diffusion model with a thermal diffusivity close to that of bulk. However, the reasonable agreement on k′ could be fortuitous. Note that the assumption of disklike morphology with a constant thickness of 3.7 nm is crude because it is likely that the distance of f-OD from the D2O/CO interface has some distribution due to heterogeneity of the ice structures in the films grown. Consequently, the small transient variations in the SFG intensities of the OD stretching bands and the lack of knowledge of the distribution of the ice-crystallite thickness do not allow us to determine whether or not the heat transport in the ultrathin ice film occurs with a thermal diffusivity similar to that of bulk. Temporal structural changes of ice on solid surfaces in response to pulsed laser heating have been studied previously. Ruan et al.58 reported in the study of ultrafast electron crystallography that the time scale for losing the hydrogen-bond network in the interfacial water on Cl/Si(111) is ∼37 ps. Kubota et al.35 reported in the study of TR-SFG that 10-layer crystalline ice on CO/Pt(111) melts in 200-400 ps. These studies showed that the randomization of the hydrogen-bond network takes place in a period longer than a few tens of picoseconds. Conversely, the time scale for propagation of the low-frequency mode excitation in the 10-layer crystalline ice of this study was estimated to be 3 ps, much shorter than those in the literature; this suggests that the low-frequency phonon excitation transfer occurs much faster than the randomization of the hydrogenbond network in crystalline ice. The heat transport and structural changes of ice on solid surfaces will be drastically altered if the pump fluence is extremely higher than that of the current study; desorption of water molecules induced by hot electron injection into the ice layer occurs within a few picoseconds at a considerably high fluence of ∼90 J m-2.59 Finally, we point out the validity of the assumption made for evaluating T j-CO(t) on the amount of heat flow from the CO monolayer to the ice overlayer. Using the temporal changes of Tw obtained above, we estimated the heat flow from the CO monolayer to the ice film in td ) 0-1 ps to be less than 10% of the photon energy absorbed by the substrate. This validates the assumption made in the previous section; i.e., the amount of heat flow at the CO/D2O interface is negligibly small. 5. Conclusions We carried out TR-SFG measurements for D2O-covered c(4 × 2)-CO/Pt(111) pumped by 150 fs pulses at λ ) 400 nm. The pump pulses excite electrons in the metal, creating hot electrons that couple with the frustrated motions of CO adsorbates and induce transient spectral changes in the CO and OD stretching regions. Analysis by simulation of the temporal responses of CO stretching bands revealed that the electron coupling time for f-CO is 1.7 ps, while that for p-CO is 1.0 ps. The difference in the electron coupling time between f-CO and p-CO implies that the 2π* state of CO is stabilized toward the Fermi level as a result of interaction with coadsorbed D2O. The excess energy

Nagao et al. transferred to the frustrated modes at the CO/D2O interface propagates into the ice layer and reaches the vacuum/ice interface within 3 ps. The time scale for the propagation of excess energy in the thin ice film is substantially shorter than those reported on laser-induced melting of ice in earlier studies. The observed dynamics were reproduced by simulations using a one-dimensional thermal diffusion model with a thermal diffusivity close to that of bulk ice. However, the lack of knowledge of the detailed structure of the ice film prevents us from clarifying heat-transport dynamics in the nanometer-scale medium, and this issue remains for future investigations. Acknowledgment. This work is supported in part by Grantsin-Aid for Scientific Research (S)(17105001) and for Young Scientists (B)(19750012) from the Japan Society for the Promotion of Science and Scientific Research on Priority Area (461 Molecular Theory for Real Systems) from the Ministry of Education, Culture, Sports, Science and Technology of Japan. K.W. gratefully acknowledges financial support by the PRESTO program of the Japan Science and Technology Agency. We thank K. Inoue for his assistance in measurements. References and Notes (1) Ueba, H. Prog. Surf. Sci. 1986, 22, 181–313. (2) Chabal, Y. J. Surf. Sci. Rep. 1988, 8, 211–357. (3) Heilweil, E. J.; Casassa, M. P.; Cavanagh, R. R.; Stephenson, J. C. Annu. ReV. Phys. Chem. 1989, 40, 143–171. (4) Cavanagh, R. R.; Heilweil, E. J.; Stephenson, J. C. Surf. Sci. 1994, 299/300, 643–655. (5) Beckerle, J. D. In Spectroscopy and Dynamics of Vibrationally Excited Adsorbates on Metal Surfaces; Dai, H.-L., Ho, W., Eds.; World Scientific: Singapore, 1995; Chapter 12, pp 459-497. (6) Ueba, H. Prog. Surf. Sci. 1997, 55, 115–179. (7) Matsumoto, Y.; Watanabe, K. Chem. ReV. 2006, 106, 4234–4260. (8) Persson, B. N. J.; Persson, M. Solid State Commun. 1980, 36, 175– 179. (9) Persson, B. N. J.; Ryberg, R. Phys. ReV. B 1985, 32, 3586. (10) Owrutsky, J. C.; Culver, J. P.; Li, M.; Kim, Y. R.; Sarisky, M. J.; Yeganeh, M. S.; Yodh, A. G.; Hochstrasser, R. M. J. Chem. Phys. 1992, 97, 4421. (11) Germer, T. A.; Stephenson, J. C.; Heilweil, E. J.; Cavanagh, R. R. Phys. ReV. Lett. 1993, 71, 3327. (12) Germer, T. A.; Stephenson, J. C.; Heilweil, E. J.; Cavanagh, R. R. J. Chem. Phys. 1993, 98, 9986–9994. (13) Germer, T. A.; Stephenson, J. C.; Heilweil, E. J.; Cavanagh, R. R. J. Chem. Phys. 1994, 101, 1704–1716. (14) Morin, M.; Jakob, P.; Levinos, N. J.; Chabal, Y. J.; Harris, A. L. J. Chem. Phys. 1992, 96, 6203–6212. (15) Katano, S.; Dobashi, S.; Kubota, J.; Onda, K.; Wada, A.; Kano, S. S.; Domen, K. Chem. Phys. Lett. 2003, 377, 601–606. (16) Bonn, M.; Hess, C.; Funk, S.; Miners, J. H.; Persson, B. N. J.; Wof, M.; Ertl, G. Phys. ReV. Lett. 2000, 84, 4653–4656. (17) Bonn, M.; Hess, C.; Wolf, M. J. Chem. Phys. 2001, 115, 7725– 7735. (18) Hess, C.; Wolf, M.; Roke, S.; Bonn, M. Surf. Sci. 2002, 502/503, 304–312. (19) Fournier, F.; Zheng, W.; Carrez, S.; Dubost, H.; Bourguignon, B. Phys. ReV. Lett. 2004, 92, 216102. (20) Fournier, F.; Zheng, W.; Carrez, S.; Dubost, H.; Bourguignon, B. J. Chem. Phys. 2004, 121, 4839–4847. (21) Backus, E. H. G.; Eichler, A.; Kleyn, A. W.; Bonn, M. Science 2005, 310, 1790–1793. (22) Symonds, J.; Arnolds, H.; King, D. J. Phys. Chem. B 2004, 108, 14311–14315. (23) Brandbyge, M.; Hedegard, P.; Heinz, T. F.; Misewich, J. A.; Newns, M. D. Phys. ReV. B 1995, 52, 6042–6056. (24) Frischkorn, C.; Wolf, M. Chem. ReV. 2006, 106, 4207–4233. (25) Ste´pa´n, K.; Gu¨dde, J.; Ho¨fer, U. Phys. ReV. Lett. 2005, 94, 236103. (26) Bonn, M.; Funk, S.; Hess, C.; Denzler, D. N.; Stempfl, C.; Scheffler, M.; Wolf, M.; Ertl, G. Science 1999, 285, 1042–1045. (27) Kizhakevariam, N.; Jiang, X.; Weaver, M. J. J. Chem. Phys. 1994, 100, 6750–6764. (28) Yamakata, A.; Uchida, T.; Kubota, J.; Osawa, M. J. Phys. Chem. B 2006, 110, 6423–6427. (29) Noguchi, H.; Okada, T.; Uosaki, K. J. Phys. Chem. B 2006, 110, 15055–15058.

Ultrafast Vibrational Energy Transfer in D2O and CO (30) Petrenko, V.; Whitworth, R. Physics of Ice; Oxford University Press: Oxford, U.K., 1999. (31) Ogasawara, H.; Yoshinobu, J.; Kawai, M. Surf. Sci. 1997, 386, 73– 77. (32) Misewich, J. A.; Heinz, T. F.; Weigand, P.; Kalamarides, A. In Laser Spectroscopy and Photochemistry on Metal Surfaces; Dai, H.-L., Ho, W., Eds.; World Scientific Publishing: Singapore, 1995; Chapter 19, pp 764-826. (33) Brandt, R. K.; Sorbello, R. S.; Greenler, R. G. Surf. Sci. 1992, 271, 605–615. (34) Kubota, J.; Wada, A.; Domen, K.; Kano, S. S. Chem. Phys. Lett. 2002, 362, 476–482. (35) Kubota, J.; Wada, A.; Kano, S. S.; Domen, K. Chem. Phys. Lett. 2003, 377, 217–222. (36) Matsumoto, Y.; Gruzdkov, Y. A.; Watanabe, K.; Sawabe, K. J. Chem. Phys. 1996, 105, 4775–4788. (37) Thiel, P.; Madey, T. Surf. Sci. Rep. 1987, 7, 211–385. (38) Henderson, M. A. Surf. Sci. Rep. 2002, 46, 1–308. (39) Richter, L. J.; Petralli-Mallow, T. P.; Stephenson, J. C. Opt. Lett. 1998, 23, 1594. (40) Klu¨nker, C.; Balden, M.; Lehwald, S.; Daum, W. Surf. Sci. 1996, 360, 104–111. (41) Su, X.; Cremer, P.; Shen, Y.; Somorjai, G. J. Am. Chem. Soc. 1997, 119, 3994–4000. (42) Bandara, A.; Katano, S.; Kubota, J.; Onda, K.; Wada, A.; Domen, K.; Hirose, C. Chem. Phys. Lett. 1998, 290, 261–267.

J. Phys. Chem. C, Vol. 113, No. 27, 2009 11719 (43) Villegas, I.; Weaver, M. J. Phys. Chem. B 1997, 101, 5842–5852. (44) Su, X.; Lianos, L.; Shen, Y. R.; Somorjai, G. A. Phys. ReV. Lett. 1998, 80, 1533. (45) Shen, Y. R. Solid State Commun. 1998, 108, 399–406. (46) Persson, B. N.; Ryberg, R. Phys. ReV. Lett. 1985, 54, 2119. (47) Persson, B. N. J.; Ryberg, R. Phys. ReV. B 1989, 40, 10273–10281. (48) Ueba, H.; Persson, B. N. J. J. Phys.: Condens. Matter 2008, 20, 224016. (49) Roke, S.; Kleyn, A. W.; Bonn, M. Surf. Sci. 2005, 593, 79–88. (50) Anisimov, S. I.; Kapeliovich, B. L.; Perel’man, T. L. SoV. Phys. JETP 1974, 39, 375. (51) Anisimov, S. I.; Rethfeld, B. Proc. SPIE 1997, 3093, 192. (52) Ueba, H.; Persson, B. N. J. Phys. ReV. B 2008, 77, 035413-5. (53) Schweizer, E.; Persson, B. N. J.; Tu¨shaus, M.; Hoge, D.; Bradshaw, A. M. Surf. Sci. 1989, 213, 49–89. (54) Baerends, E.; Ros, P. Mol. Phys. 1975, 30, 1735–1747. (55) Joshi, A. A.; Majumdar, A. J. Appl. Phys. 1993, 74, 31–39. (56) Slack, G. A. Phys. ReV. B 1980, 22, 3065. (57) Tombari, E.; Salvetti, G.; Ferrari, C.; Johari, G. P. J. Chem. Phys. 2005, 123, 214706-5. (58) Ruan, C.-Y.; Lobastov, V. A.; Vigliotti, F.; Chen, S.; Zewail, A. H. Science 2004, 304, 80–84. (59) Backus, E. H. G.; Grecea, M. L.; Kleyn, A. W.; Bonn, M. J. Phys. Chem. B 2007, 111, 6141–6145.

JP901793Q