Wavelength Dependence of UV Photoemission from Solvated

---

Article pubs.acs.org/JPCA

Wavelength Dependence of UV Photoemission from Solvated Electrons in Bulk Water, Methanol, and Ethanol Yo-ichi Yamamoto, Shutaro Karashima, Shunsuke Adachi, and Toshinori Suzuki* Department of Chemistry, Graduate School of Science, Kyoto University, Kitashirakawa-Oiwakecho, Sakyo-Ku, Kyoto 606-8502, Japan S Supporting Information *

ABSTRACT: We have measured the wavelength dependence (340−215 nm) of one-photon photoemission from the ground electronic state of solvated electrons in bulk water, methanol, and ethanol. In every case, the vertical electron binding energy (VBE) gradually increased with photon energy, indicating that the photoelectron kinetic energy diminishes as a result of electron−vibration inelastic scattering prior to emission from the liquid surface. In contrast, the VBE of the Rydberg electron in DABCO (1,4-diazabicyclo[2,2,2]octane), which has a surface-excess density, revealed no clear wavelength dependence. These results suggest that the solvated electrons are created predominantly in the bulk and that VBEs measured using UV photoemission spectroscopy of liquids generally require energy corrections to account for inelastic scattering effects. From the wavelength dependence, we have re-estimated the VBEs of solvated electrons in bulk water, methanol, and ethanol to be 3.3, 3.1, and 3.1 eV, respectively. Hydrated electrons were also identified by photoemission spectroscopy using 90 nm radiation.

I. INTRODUCTION The water in a living cell exposed to radiation undergoes ionization to produce OH radicals and hydrated electrons [e−(aq)]. The structure and dynamics of e−(aq) are thus of great interest in radiation chemistry and biology.1 The simplest physical model for describing e−(aq) is an electron trapped in a spherical dielectric cavity. Although this conventional picture has been challenged,2 recent quantum chemical studies using DFT (density functional theory) seem to suggest that e−(aq) indeed exists in a cavity state, although a considerable fraction of the excess electron density is distributed over multiple hydration shells.3−5 At 300 K, an excess electron attracts, on average, four water molecules in the first hydration shell by orienting one of two OH bonds of each molecule toward the center of the e−(aq).4,5 The radius of gyration of e−(aq) has been calculated to be around 0.25 nm.1,6,7 One of the most fundamental properties of e−(aq) relevant to its chemical reactivity is the electron binding energy (eBE). Since the hydration structure of e−(aq) is completely unfavorable for the hydrogen-bonding of neutral water molecules without an excess electron, photoemission of e−(aq) is an electronic transition to a repulsive part of the intermolecular potential of liquid water. Therefore, photoemission spectroscopy of e−(aq) provides not an adiabatic eBE, but a “vertical” electron binding energy (VBE). In 1990, Coe et al. extrapolated the VBE of negatively charged water clusters, (H2O)−n (n = 2−69) and estimated the VBE of e−(aq) to be 3.3 eV.8−10 Subsequently, Verlet et al. found that (H2O)−n exists in © XXXX American Chemical Society

three isomer series with different VBEs, and they suggested that series II and III contain a surface-bound excess electron, while series I, used by Coe et al. for their prediction of VBE of e−(aq),8 contains an excess electron in the cavity state.11 However, Turi et al. argued that the excess electron is surfacebound for all negatively charged water clusters.12−14 Hammer et al. also reported infrared spectroscopy of small clusters of series I, which revealed a signature of a double-acceptor water molecule binding an excess electron with free OH bonds on the surface.15 Ma et al. revisited photoemission spectroscopy of negatively charged water clusters by annealing and cooling to 10 K in a cryogenic ion trap, and found that the series I clusters can be further categorized into two subseries, denoted Ia and Ib.16 The VBE of e−(aq) was estimated to be around 4 eV using the subseries Ib, which becomes the dominant species at n = 46,16 although the electronic and geometrical structures of the Ib clusters have yet to be established. In 2010, the VBE of e−(aq) was directly measured for the first time by three independent groups using photoemission spectroscopy of liquid microjets.17−20 The reported values, from 3.3 ± 0.1 to 3.6 eV, were in good agreement with each other, despite the fact that some of the earlier experiments were limited by low signal-to-noise ratios and/or the lack of energy calibration against streaming potentials.21−23 The VBE value Received: October 1, 2015 Revised: December 7, 2015

A

DOI: 10.1021/acs.jpca.5b09601 J. Phys. Chem. A XXXX, XXX, XXX−XXX

Article

The Journal of Physical Chemistry A

case, how accurate is the estimate of 3.4 eV for the VBE of e−(aq)? In this paper, we show that the estimate of VBE for e−(aq) by UV photoemission spectroscopy of liquid microjets varies with the probe wavelength, which is likely due to the influence of inelastic scattering. Vacuum UV photoemission spectroscopy of e−(aq) using 90 nm radiation is also performed for the first time.

was further refined to 3.4 eV following a high-resolution photoemission spectroscopy study of e−(aq) using a linear time-of-flight (TOF) spectrometer with a 100 kHz femtosecond laser.24 Photoemission spectroscopy is generally considered to be a surface-sensitive method, while its actual probing depth is unknown at low electron kinetic energy (eKE) for any material, including liquids. Therefore, it is natural to ask whether the VBEs of e−(aq) measured using UV photoemission spectroscopy of liquids can be considered as the bulk values. Two research groups25−27 have estimated the effective attenuation length (EAL) of an electron flux in liquid water using soft X-ray photoemission spectroscopy, although these measurements were limited to eKE > 10 eV, and the EAL values differed by a factor of 2−3 from each other. When the EAL values by Suzuki et al. are extrapolated, the EAL for eKE < 10 eV is estimated to be at least several nanometers.27 On the other hand, extrapolation of the values measured by Thürmer et al. suggests a subnanometer EAL in the same region.26 Additionally, UV photoemission spectroscopy of surface-active solutes by Lübcke et al. provided an estimated EAL of 5 nm at an eKE of 3 eV.28 On the basis of these measurements, it is likely that the probing depths in the UV photoemission spectroscopy of liquids for eKE < 10 eV are on the order of several nanometers, which is more than an order of magnitude greater than the radius of gyration of e−(aq). Therefore, in this probing depth, the properties of e−(aq) are expected to essentially match those in the bulk. Most of the VBE measurements for e−(aq) have employed charge-transfer-to-solvent (CTTS) reactions from I−,29 for which molecular dynamics simulations have predicted a surfaceexcess density;30 therefore, creation of a surface-bound electron could not be readily ruled out. However, the observed signal at a VBE of 3.4 eV has been assigned to e−(aq) in the bulk for the following reasons. First, liquid water at room temperature has sufficient internal energy and structural fluctuations to relax an excess electron into a cavity state, so that the lifetime, if any, of a surface-bound state of e−(aq) is expected to be short.31 Since the VBE measurements were performed at time delays of several hundred picoseconds or more after the generation of e−(aq), any surface-bound states32 should have been quenched. Second, photoemission spectroscopy via the CTTS state of I− does not exhibit any anisotropy,33,34 suggesting that I− is well hydrated within the probing depth of photoemission spectroscopy and that a photoelectron from I− is elastically scattered in the bulk prior to emission from the surface.35,36 Therefore, e−(aq) created by CTTS from I− is also expected to be well hydrated. Moreover, photoemission from the Rydberg state of DABCO (1,4-diazabicyclo[2,2,2]octane), which has a surfaceexcess density, exhibits clear anisotropy, while e−(aq) produced from DABCO exhibits isotropic photoemission.34 This indicates that, even if CTTS occurs from a surface-active species, e−(aq) is created on the bulk side.34 Our computational simulation using the CMS-Xα method showed that photoemission from DABCO and e−(aq) are expected to be isotropic when they have complete hydration shells.34 Thus, the VBE value of 3.4 eV was assigned to e−(aq) in the bulk or to fully hydrated e−(aq) near the surface. Given that UV photoemission from e−(aq) is subject to scattering in solution, one may wonder, if there is the influence of inelastic scattering on the estimated eBE. When inelastic scattering is significant, the eBEs measured by UV photoemission spectroscopy require an energy correction. In that

II. EXPERIMENTAL SECTION II.A. UV Photoemission Spectroscopy. A 1 kHz Ti:sapphire regenerative amplifier (100 fs, 2.3 mJ/pulse) was employed to excite two optical parametric amplifiers (OPAs), and the outputs from the OPAs were used directly as the pump and probe pulses in the wavelength region between 400 and 240 nm. The wavelength region below 240 nm was covered using frequency mixing of the output from one OPA with the Ti:sapphire fundamental (800 nm) in a BBO crystal. The crosscorrelation time between the pump and probe pulses was typically 250 fs. The probe pulse was optically delayed with respect to the pump pulse using a computer-controlled linear stage, and both the pump and probe pulses were focused on a liquid microjet. Typical laser pulse energies of the pump and probe pulses were 5 and 10 nJ/pulse, respectively. A liquid laminar flow was discharged from a fused silica capillary with a 15 μm inner diameter into a vacuum chamber at a flow rate of 0.2 mL/min.37 The concentration of the solutes and solvents and the pump and probe wavelengths employed in this study are listed in Table 1. The liquid microjet was Table 1. Experimental Conditions for the VBE Measurements of This Study solute and concentration

solvent

pump wavelength (nm)

probe wavelength (nm) 240−340 same as pump 240−340 same as pump 240−340 same as pump 260−400 90

0.1 M NaI

water

0.1 M NaI

methanol

0.1 M NaI

ethanol

0.2 M DABCO, 0.1 M NaCl 0.5 M K4[Fe(CN)6]

water

226 215−230 226 215−230 226 215−230 226

water

270

photoexcited with the pump pulse 1 mm downstream from the capillary nozzle and photoemission was induced with the probe pulse. The photoelectrons emitted from the liquid surface were sampled using a skimmer with an entrance diameter of 0.5 mm located 2 mm from the liquid surface. Photoelectrons were collected and analyzed using a magnetic bottle photoelectron spectrometer, which consists of a SmCo permanent magnet with a conical iron cap and a 1.2 m flight tube with a solenoid and a permalloy shield. The photoelectrons were detected using a dual microchannel plate placed at the end of the flight tube, and the photoelectron signal was preamplified and counted using a multichannel scaler. The linear polarization of the pump laser was set parallel to the flight axis, and the linear polarization of the probe laser was either set parallel or perpendicular to the flight axis for 215−300 nm and 320−400 nm, respectively. The electron detection efficiency of a magnetic bottle photoelectron spectrometer generally decreases at low eKE. Therefore, the observed eKE distribution needs to be corrected B

DOI: 10.1021/acs.jpca.5b09601 J. Phys. Chem. A XXXX, XXX, XXX−XXX

Article

The Journal of Physical Chemistry A

distribution is normalized at its peak. Hydrated electrons were produced by CTTS from I− by the pump pulse, and a pump− probe time delay of 600 ps was employed to thermalize e−(aq) prior to photoemission. As expected, we observed an increase in eKE with hνprobe; however, while the high-energy edge of the eKE distribution shifts by exactly the amount of increase in hνprobe, the peak shifts by only a fraction of it. Consequently, the width of the eKE distribution gradually increases with hνprobe. Using spline fitting of the observed distributions, a 2D map of the eBE distribution vs hνprobe was constructed, as shown in Figure 2. The map clearly indicates that VBE, defined

for the energy-dependent transmission efficiency of the spectrometer, or the measurements should be performed in the energy region where the transmission is constant. We have measured the transmission efficiency of our magnetic bottle spectrometer using [1 + 1] resonance enhanced multiphoton ionization of NO at various pass energies and confirmed that electrons with a pass energy higher than 0.3 eV are transmitted perfectly. Therefore, we biased the field-free region of our spectrometer to +0.5 V to increase the electron pass energy to the region where transmission is uniform. The influence of the bias voltage on the eKE has been accounted for in the analysis. II.B. VUV Photoemission Spectroscopy. The output from a home-built Ti:sapphire amplifier (810 nm, 12 mJ, 35 fs at 1 kHz) was split into two branches, each of which was converted to the third harmonic (3ω).38 One of these 3ω beams was used as the pump. The time duration of the pump pulses were intentionally stretched using a synthetic quartz crystal to minimize one-color multiphoton photoemission from the sample while achieving a high excitation efficiency. The other 3ω beam was loosely focused into a gas flow cell filled with Kr to generate the ninth harmonic (9ω), which was used as the probe. The generated harmonic was reflected twice by SiC Brewster-angled beam splitters to separate the 3ω beam from the 9ω beam. The incident angle with the SiC beam splitters was set to 71.8° (Brewster angle for the 3ω). The cross-correlation time between the pump and probe pulses was 1 ps, and the laser pulse energy of the pump beam was 1 μJ/ pulse. The magnetic bottle photoelectron spectrometer used in the VUV studies was almost the same as that used in the UV experiments described in the previous section, but the transmission efficiencies of the two are different. In the VUV experiment, we applied a DC voltage of −4 V to the field-free region to reject slow electrons with an eKE of less than 4 eV, as well as to lower the transmission efficiency of electrons with an eKE of less than 5 eV. The eKE values recorded while applying the bias voltage have been corrected in the analysis.

Figure 2. Electron binding energy (eBE) distribution for hydrated electrons as a function of hνprobe. The map was constructed using third-order spline fitting of the spectra shown in Figure 1. The cross symbols (+) indicate the peak positions for the 11 eBE distributions observed at different hνprobe.

as the peak position of the eBE distribution, increases with hνprobe. The cross symbols used in the map are the peak positions for the 11 distributions shown in Figure 1. Figure 3 presents similar 2D maps for solvated electrons in (a) methanol and (b) ethanol calculated from the eBE distributions presented in Figures S1 and S2 of the Supporting Information, respectively; in both cases, VBE increases with hνprobe. These results indicate that the solvated electrons are created in the bulk, and that inelastic scattering between a photoelectron and solvent (or solute) molecules reduce eKE prior to electron emission from the surface. If the wavelength dependence of the eBE distribution arises from inelastic scattering in the bulk, a similar dependence will not occur for photoemission from chemical species on the liquid surface. Therefore, we performed a similar experiment for an aqueous DABCO solution, for which a surface-excess solute density has been confirmed by soft X-ray photoemission spectroscopy.34 In our previous study, angle-resolved photoemission spectroscopy of an aqueous DABCO solution revealed anisotropic photoemission from the Rydberg state of DABCO, which indicated that elastic scattering is not significant in photoemission from DABCO.34 In this study, we measured the eBE distribution for photoemission from the Rydberg state of DABCO at a pump−probe time delay of 200 fs and also subtracted the small contribution from e−(aq). As shown in Figure S3 of the Supporting Information, the pump wavelength of 226 nm falls on the shoulder of the photoabsorption spectrum of aqueous DABCO solution, for which the upper state is the 3p and/or 3s Rydberg state. The results shown in Figure 4 clearly indicate that the VBE for DABCO does not vary with hνprobe, except at very low hνprobe,

III. RESULTS AND DISCUSSION A. Electron Binding Energies for Solvated Electrons Measured as a Function of hνprobe. Figure 1 shows onephoton photoemission spectra of e−(aq) measured at 11 different probe photon energies (hνprobe); the intensity of each

Figure 1. Electron kinetic energy (eKE) distributions in ultraviolet one-photon photoemission of hydrated electrons produced by CTTS from I− in aqueous NaI solution (see Table 1 for the experimental conditions). The eKE at the peak and the fwhm of each distribution are listed in Table S1 of the Supporting Information. C

DOI: 10.1021/acs.jpca.5b09601 J. Phys. Chem. A XXXX, XXX, XXX−XXX

Article

The Journal of Physical Chemistry A

previous measurements were performed at hνprobe of 260 nm. The present study is the first systematic investigation of VBE at multiple wavelengths, and we have observed an unambiguous wavelength dependence. Previously, Shreve et al. reported that the VBE of e−(aq) was insensitive to the photoemission wavelength, choice of parent anion, and anion concentration.20 They employed two wavelengths of 266 and 213 nm, which created mean eKEs of 1.1 and 2.2 eV in the photoemission from e−(aq), respectively. The 266 nm experiment used aqueous [Fe(CN)6]4− solutions in the concentration range of 50−250 mM, while the 213 nm experiment used both an aqueous 50−100 mM [Fe(CN)6]4− solution and an aqueous 100 mM I− solution. In every measurement, they obtained a VBE in the range of 3.5−3.7 eV.20 We performed similar experiments on an aqueous 100 mM [Fe(CN)6]4− solution and observed a wavelength dependence similar to that observed in Figure 2. It is unclear why Shreve et al. observed no wavelength dependence; however, [Fe(CN)6]4− experiments at short wavelengths suffer from interference due to strong one-photon photodetachment from [Fe(CN)6]4−, even below its VBE of 6.1 eV, which leads to some complications in the data analysis. From the VBE values measured at the longest probe wavelength, we conclude that the best estimates of the VBE for solvated electrons in water, methanol, and ethanol are 3.3, 3.1, and 3.1 eV, respectively. It is important to note that these values are for the 0.1 M solutions employed in this study. Our preliminary results indicate that the VBE values vary slightly with the solution concentration, although the variation is less than 0.3 eV for NaI solutions. B. Vacuum UV Photoemission Spectroscopy of e−(aq). As described in the previous section, the apparent VBE of e−(aq) increases continuously with hνprobe up to 5.8 eV. For technical reasons, it was difficult for us to perform similar experiments at hνprobe > 5.8 eV. Therefore, we have instead employed the ninth harmonic (9ω: hνprobe = 13.6 eV) of a Ti:sapphire laser. To date, the only VBE measurement of e−(aq) performed at a photon energy higher than the ionization energy of solvent water (11.3 eV) was carried out by Siefermann et al. (38.7 eV).17 It was therefore intriguing to perform another VUV photoemission experiment at 13.6 eV; however, as described below, the experiment turned out to be rather difficult and complex. Figure 5 shows a portion of the photoemission spectrum of aqueous 0.5 M K4[Fe(CN)6] solution measured using 9ω. The broad feature seen in the spectrum is the 1b1 [HOMO−1] band of the solvent water, while the narrow band is that of water vapor that had evaporated from the liquid surface. The 1b1 binding energies of gaseous and liquid water have been reported to be 12.6 and 11.3 eV,23 respectively, while the energy difference seen in Figure 5 is only 0.8 eV. The 9ω (13.6 eV) photon causes one-photon ionization of liquid water, which produces electrons with an eKE of 2.3 eV. We have already seen that for photoemission of e−(aq), the eKE in this energy range is altered by inelastic scattering in solution (Figures 1 and 2). A reduction in eKE will increase the apparent eBE of liquid water to reduce the difference from eBE of gaseous water. Thus, the results suggest that eBE measured for liquid water should be adjusted by −0.5 eV to correct for inelastic scattering. The full width at half-maximum of 1b1(L) is 1.78 eV, which is greater than the value of 1.45 eV previously obtained from soft X-ray photoemission spectroscopy. The larger width is also indicative of inelastic scattering.

Figure 3. Electron binding energy (eBE) distribution for solvated electron in (a) methanol and (b) ethanol as a function of hνprobe. The map was constructed using third-order spline fitting of the photoemission spectra (not shown here) similar to Figure 1. The cross symbols (+) indicate the peak positions for the 11 eBE distributions observed at different hνprobe.

Figure 4. Electron binding energy (eBE) distribution for photoemission from the Rydberg state of DABCO in aqueous solution as a function of hνprobe. The map was constructed using third-order spline fitting of the photoemission spectra (not shown), similar to Figure 1. The small contribution from photoemission of hydrated electrons has been subtracted. The pump−probe time delay was 200 fs. The cross symbols (+) indicate the peak positions for the 9 eBE distributions observed at different hνprobe. It should be noted that the energy region of hνprobe in this figure is lower than that in Figures 2 and 3.

where the photon energy is insufficient to cover the Franck− Condon envelope of photoionization of DABCO. Therefore, the lack of wavelength dependence for DABCO agrees with our expectation. Photoemission from an aqueous DABCO solution after 3 ps is predominantly from hydrated electrons in the bulk,34 and we confirmed that photoemission after 3 ps exhibits a wavelength dependence similar to that seen in Figure 2. Table 2 summarizes the experimental conditions employed in previous VBE measurements for e−(aq). Most of the D

DOI: 10.1021/acs.jpca.5b09601 J. Phys. Chem. A XXXX, XXX, XXX−XXX

Article

The Journal of Physical Chemistry A

Table 2. Vertical Electron Binding Energies (eV) Previously Reported for Solvated Electrons in Three Solvents, and the Probe Laser Wavelength (nm) and Precursors Employed in Those Measurements Siefermann17 water methanol ethanol probe wavelength (nm) precursor

3.3

Tang18 3.3

Lübcke19 3.4

Shreve20 3.5−3.7

32.1

260

266

266, 213

0.5 M K4[Fe(CN)6]

0.14 M NaI

0.1 M NaI

0.05−0.25 M K4[Fe(CN)6], KI

Figure 5. Photoemission spectrum of liquid water measured using 9ω (90 nm). The strong narrow band is the 1b1 band of gaseous water, and a broad band is the 1b1 band of liquid water.

Horio24

Shen39

Shreve40

3.4 3.4 3.3 260

3.1 3.1 260

3.4 3.4 213, 231

0.1 M NaI

0.14 M NaI

0.1 M KI, 0.2 M NaI

Figure 6. Semilog plot of the photoemission spectra of aqueous K4[Fe(CN)6] solution measured for different pump (3ω)−probe (9ω) time delays. The two spectra were integrated over the delay times of −18 to 0 (red) and 0 to 18 ps (blue). The inset shows the difference spectrum plotted on a linear scale, and corresponds to the photoemission signal from e−(aq).

In the VUV photoemission spectroscopy experiments, the signal from e−(aq) was weaker than the solvent water signal by 6−7 orders of magnitude; therefore, we employed relatively high probe pulse intensities to try to increase the small signal. Consequently, a space-charge was present in our ionization volume. (As described in the Experimental section, we have intentionally stretched the pump pulse duration to minimize multiphoton photoemission induced by the pump pulse.) Space-charge effects are a common problem in time-resolved pump−probe photoemission spectroscopy in the VUV and Xray regions when studying minor constituents in condensed matter, such as solutes, impurities, and adsorbates.41,42 We found that the eBEs were shifted by −0.3 eV at high probe pulse intensities. With an energy correction of −0.3 eV, the band position of [Fe(CN)6]4− was in fair agreement with the literature value of 6.1 eV.43 The spectrum was rather broad, which may be partly due to the formation of [Fe(CN)6]3− (VBE = 7.6 eV), as reported by Siefermann et al.17 Figure 6 shows a semilog plot of the photoemission spectra in the lowest eBE region with and without the pump pulse. Both spectra were integrated for the time delays of −18 to 0 (red) and 0 to 18 ps (blue) because the signal levels were so low. The difference spectrum between these two spectra plotted on a linear scale is presented in the inset, and corresponds to the signal from e−(aq). In principle, the spectrum involves a contribution from the CTTS state of [Fe(CN)6]4−; however, the lifetime of the CTTS state is very short (