Working on a dream: bringing up the level of interface spectroscopy to the bulk level

Abstract

Liquid interfaces are unique environments in which a variety of fundamental phenomena occur. Therefore, it is important to obtain a molecular-level understanding of liquid interfaces for both basic science and industrial applications. However, it is not an easy task to investigate molecules in the interface region that only has nanometer thicknesses. Second-order nonlinear spectroscopy, or even-order nonlinear spectroscopy more generally, is intrinsically interface-selective because the relevant nonlinear signal is generated only in the region in which the inversion symmetry is broken under the dipole approximation. In the past 2 decades, we have been developing and applying new interface nonlinear spectroscopic methods, aiming to bring up the level of knowledge on interfacial molecules to that on molecules in solution. During this attempt, we developed electronic sum-frequency generation spectroscopy, heterodyne-detected electronic sum-frequency generation spectroscopy, and heterodyne-detected vibrational sum-frequency generation spectroscopy, as well as fourth-order Raman spectroscopy. We also extended the methods to femtosecond time-resolved measurements. Using these methods, we are now able to study the structure and dynamics at liquid interfaces, in particular exposed interfaces such as air/liquid interfaces, at a similar level to the study for solution. I overview our interface research while describing thoughts we had at each turning point.

Graphical Abstract

Liquid interfaces are unique environments in which a variety of fundamental phenomena take place. In the past 2 decades, we have been developing and applying new interface nonlinear spectroscopic methods, aiming to bring up the level of knowledge on interfacial molecules to that on molecules in solution. I overview our interface research, focusing on the thoughts we had at each point of the research.

Open in new tabDownload slide

1. Introduction

Liquid interfaces are unique environments in which a variety of fundamental molecular phenomena take place. Therefore, it is critically important to obtain a molecular-level understanding of liquid interfaces not only for basic science, but also for industrial applications. However, it is not an easy task to investigate molecular processes occurring in very thin interface regions that are of only nanometer thickness. Second-order nonlinear spectroscopy, or even-order nonlinear spectroscopy more generally, is intrinsically interface-selective because the relevant nonlinear signal is generated only in the region in which the inversion symmetry is broken under the dipole approximation. In the past 2 decades, we have been developing and applying new interface nonlinear spectroscopic methods, aiming to bring up the level of our knowledge on interfacial molecules to that on molecules in solution. Our first paper on interface research was published in 2004,¹ although I started thinking of this new research direction a little earlier than that. Looking back on the long-term efforts that we have made, I might be allowed to say that we have made some progress in understanding liquid interfaces, in particular for aqueous interfaces. Because I was fortunately provided with an opportunity to write an Award Account in the Bulletin of the Chemical Society of Japan, I thought that this would be a good chance for me to write about how we have proceeded with interface research, including the progress after I received the CSJ Award for Creative Work in 2012, for young researchers who are willing to take on challenges to start something new. We have already published several review articles about each of the new interface-selective nonlinear spectroscopic methods that we developed, i.e. electronic sum-frequency generation (ESFG),² heterodyne-detected vibrational sum-frequency generation (HD-VSFG),³ and time-resolved HD-VSFG (TR-HD-VSFG) as well as two-dimensional HD-VSFG (2D HD-VSFG).⁴ Therefore, I will write this article by mainly focusing on how we thought and made decisions at each point of our interface research. Because of this aim, I do not intend to provide a full set of references, but rather to focus on describing the stream of our own research. I would like readers to see our original papers and reviews if they need to find the full set of references about each research subject that I discuss in this article.

2. Preparation

In 2001, I was appointed Chief Scientist at RIKEN and was given the chance to start a new laboratory from scratch. RIKEN is a research institute that has a long history that started in 1917 and it had expanded to become the largest research institute in Japan when I joined. A unique system in RIKEN is the Chief Scientist system, in which a newly appointed Chief Scientist is allowed to launch a new laboratory by appointing typically 3 permanent researchers with complete freedom to choose the research subjects and directions. Before joining RIKEN, I enjoyed running a relatively small research group consisting of about 5 people at the Institute for Molecular Science (IMS) in Okazaki, Japan, to carry out research on ultrafast spectroscopy as an associate professor. Based on this experience, I thought that it would be a good idea to make 3 subgroups carrying out research on different subjects in the new Chief Scientist Laboratory with 3 staff scientists I could appoint. It meant that I needed to generate new ideas about what kind of challenges I should take on with each of them. Although I only had vague images about future research directions at that time, I was young and I wanted to try many things anyway.

I had already decided to continue research on ultrafast spectroscopy at RIKEN, which I started at the IMS. So, I appointed my coworker at the IMS, Dr. Satoshi Takeuchi, as the first staff scientist in my laboratory at RIKEN to launch a subgroup for ultrafast spectroscopy, with a particular focus on the observation of the nuclear wave-packet motion of reacting molecules.^5–10 For the other subjects, I was initially not sure which subjects deserved new challenges. Nevertheless, I had already decided to invite Dr. Shoichi Yamaguchi, who was working in a company at that time, to my new laboratory at RIKEN as the second staff scientist. I met Shoichi for the first time when I worked as a research associate in a 5-yr research project headed by Prof. Hiro-o Hamaguchi at Kanagawa Academy of Science and Technology (KAST). Shoichi joined this “maverick” project immediately after obtaining his master's degree in the Department of Physics at the University of Tokyo and we worked together as colleagues for a while. During this period, I sensed his excellence in experiments and very pure mind for science. So, I wanted to start new research with him at RIKEN. Even though I had firmly made up my mind to appoint him as the second staff scientist and Shoichi happily agreed to join my laboratory privately, it took about a year to complete all the official processes to prepare for his position at RIKEN. Because we were so eager to start new research that we could not wait without doing anything, we decided to meet once a month at a restaurant in the Shinjuku area of Tokyo to discuss the research subject that we should pick up while eating dumplings with beer. On reflection, no good concrete idea emerged in these secret meetings, but it was truly precious time for me to incubate ideas about new research. It was one of the most creative times in my research life so far.

Shoichi joined RIKEN in 2002, but still I was not very sure which research direction we should head in. Nevertheless, slowly, slowly, 2 ideas came to my mind. One of them was the interfaces. It was the true start of a series of our research that I will describe in this article. The other idea was the study of asynchronous dynamics that we cannot investigate using the pump-probe method. I started research on this idea later with the third staff scientist, Dr. Kunihiko Ishii, and we developed new single-molecule spectroscopy having sub-microsecond time resolution, which we named two-dimensional fluorescence lifetime correlation spectroscopy (2D FLCS).^11–14

Since I was trained as a Raman spectroscopist under the supervision of Profs. Mitsuo Tasumi and Hiro-o Hamaguchi in graduate school at the University of Tokyo, I knew the work of Prof. Y. R. Shen on interface-selective second-order nonlinear spectroscopy from very early on.^15–18 In fact, I wrote a short Japanese article in Kagaku-to-Kogyo (a monthly magazine published by the Chemical Society of Japan) when I was a graduate student and introduced the earliest studies about interface-selective VSFG by the Shen group.¹⁹ Since then, I had been a good “audience” of interface-selective nonlinear spectroscopy for a long time. So, I knew well that many research groups were intensively utilizing the VSFG spectroscopy that the Shen group had developed and that they reported vibrational spectra of various interfaces. In this sense, the usefulness of VSFG spectroscopy was already established in the interface community before I started interface research. However, from the “naive” viewpoint of a scientist who was studying ultrafast dynamics using femtosecond time-resolved spectroscopy in solution, interface-selective nonlinear spectroscopy at that time still looked immature compared with the variety of advanced spectroscopy that we utilized for investigating molecules in solution. So, I thought that, if we could bring up the level of interface-selective nonlinear spectroscopy to the level of the most advanced spectroscopy for molecules in solution, we might be able to deepen our molecular-level understanding of the interface while attempting to observe intriguing fundamental interfacial phenomena. This thought made me decide to start the research on liquid interfaces by developing new types of interface-selective nonlinear spectroscopy.

3. Start: ESFG and its application

To start new research, a concrete practical idea is indispensable. I strongly believe that, in general, it is dangerous to enter an unfamiliar research field just because it looks popular. Therefore, when I decided to start interface-selective nonlinear spectroscopy, I also decided not to start it with VSFG because VSFG spectroscopy was already widely utilized for investigating a variety of interfaces such as liquid and polymer interfaces. Actually, it looked to me that there was very little chance for us to do something new if we started with such a well-established technique by just following other active research groups. Thus, I needed a new idea to start our own interface study.

In 2003, I had a chance to attend an international conference held in Singapore. Although I was invited to give a talk about ultrafast spectroscopy,²⁰ I tried to hear talks about as many different subjects as possible, seeking clues for our new research. Then, I happened to hear a talk about an interface study carried out using second harmonic generation (SHG) spectroscopy. They reported electronic spectra of interfacial molecules that were obtained by plotting the SHG intensities measured by changing the wavelength of the ultrashort excitation laser over a wide wavelength region. It was a good talk, but the quality of the spectra shown was not good because their electronic spectra consisted of only a very small number of data points. While listening to the talk, somehow an idea came to my mind: they use one single-color light for measuring the second-order electronic response, but what if we use 2 differently colored lights? Then, if we use broadband light for one light, would it possible to obtain interface-selective electronic spectra in a wide wavelength region without scanning the color of the light? I immediately realized that this idea was technically very feasible because we were routinely using the white-light continuum as the probe for femtosecond time-resolved absorption experiments in our laboratory. In other words, I was very sure that we could obtain interface-selective electronic spectra with a single experiment by measuring the sum frequency of the narrowband output of a Ti:sapphire amplifier and the white-light continuum. The critically important point for me was that nobody had done it before, although the idea was simple. I went back to RIKEN and explained this idea to Shoichi with excitement. He quickly understood the meaning of it and, within a very short time, demonstrated that this method can certainly provide high-quality interface-selective electronic spectra with a much shorter measurement time compared with the scanning SHG method. We published our first paper on interface-selective nonlinear spectroscopy and reported this technique.¹ We named it electronic sum-frequency generation (ESFG) spectroscopy.²¹

Figure 1a shows a typical ESFG spectrum that was obtained from a dye molecule, coumarin 338 (C338), at the air/water interface.²² Two pulses were used for the ESFG measurement: one was the output of a Ti:sapphire amplifier at 795 nm (ω₁, bandwidth ∼160 cm^–1) and the other was the white-light continuum (ω₂, 540 nm∼1.2 μm), which was generated by focusing the output of the Ti:sapphire amplifier into water. The vertical axis of the spectrum stands for the normalized signal intensity, which was obtained by normalizing the observed sum-frequency signal intensity with that of a quartz crystal. The horizontal axis represents the wavelength corresponding to ω₁ + ω₂. For C338, the two-photon resonant excited state corresponds to the S₁ state and there is no electronic state that is one-photon resonant with ω₁ or ω₂. Therefore, the ESFG spectrum plotted along ω₁ + ω₂ directly represents the electronic structure of C338 at the air/water interface. As shown in the figure, ESFG spectroscopy provides an interface-selective electronic spectrum with very high quality that is equivalent to that of ultraviolet (UV)-visible absorption spectra in solution (Fig. 1b). I also note that coumarin dyes are known to exhibit solvatochromic shifts: the energy of the electronic transition changes with the change in the solvent polarity. In fact, the peak position of the electronic spectrum of C338 in hexane appears at ∼400 nm whereas that in water appears at ∼445 nm. Importantly, the peak of the ESFG spectrum at the air/water interface is located in between the peaks in nonpolar hexane and polar water, indicating that C338 experiences a middle polarity at the air/water interface. This observation accords with our intuition for a solute molecule at the air/water interface, i.e. only half of the molecule is solvated by polar water and the other half is exposed to the nonpolar air side. This spectrum clearly manifests that the ESFG selectively measures the electronic spectrum of C338 at the air/water interface with a high S/N equivalent to that of the absorption spectra in solution.

The ESFG spectra of solvatochromic coumarin dyes opened the door to quantitatively consider the polarity at the water interface. The polarity of the environment is a very essential factor that controls the properties and reactivities of molecules, and hence it is a very important quantity in chemistry. On the analogy of the bulk solvent polarity, the effective polarity of liquid interfaces had been discussed as a scale that expresses the ability of solvent molecules to stabilize polar solute molecules at the interface. In the pioneering work of Eisenthal and coworkers, they measured the electronic spectra of solvatochromic molecules at interfaces by using SHG and proposed that the effective polarity of the liquid interface was given by the arithmetic average of the polarities of the 2 constituent bulk phases.^23,24 Sobhan Sen picked up this problem in our laboratory and quantitatively discussed the polarity of liquid interfaces by measuring the electronic spectra of solvatochromic coumarin molecules using ESFG that provides electronic spectra at the interface with unprecedentedly high quality.²⁵

Figure 2a shows the ESFG spectra of 5 coumarin dyes at the air/water interface, which are compared with UV-visible absorption spectra in various solvents.²⁵ The peak wavelength of the ESFG spectra of all the coumarin dyes at the air/water interface is located in between the peaks in water and nonpolar hexane, as previously reported with the low S/N interfacial electronic spectra measured by the scanning SHG method.^23,24 Surprisingly, however, the high-quality ESFG spectra revealed that the effective polarities of the air/water interface indicated by different coumarin dyes were significantly different from each other. For example, the ESFG spectrum of coumarin 314 (C-314) exhibits almost the same peak wavelength as that in bulk butyl ether, whereas the spectrum of coumarin 110 (C-110) at the interface shows almost the same peak wavelength as that in bulk ethanol. This implies that different coumarin molecules experience different polarities at the same air/water interface. In Fig. 2b, the peak wavelengths in different solvents and at the air/water interface are plotted against the normalized E_T(30) polarity scale (E_T^N) of the solvents for each coumarin. This plot clearly shows that the local solvation environments around different coumarin molecules are different at the air/water interface, although they have similar molecular structures. This result initially puzzled us, but we found that it could be rationalized in terms of a small difference in the tilt angle and vertical position of the solute at the water interface, which causes a substantial difference in the stabilization energy that the solute gains from solvation at the interface. This study taught us that the effective polarity cannot be defined as an intrinsic property of the air/water interface, but that it substantially changes, depending on the solute molecules. Through this work, I was convinced that liquid interfaces have a lot of unexplored fundamental problems and that they can be studied by developing new interface-selective nonlinear spectroscopy.

Based on the ESFG spectroscopy that we first developed, we also realized frequency-domain fourth-order nonlinear (χ⁽⁴⁾ Raman) spectroscopy that provides interface-selective vibrational spectra.^26,27 Although χ⁽⁴⁾ Raman spectroscopy is technically demanding, it has high potential for the study of buried interfaces because it can be performed only with visible light and can hence reach the interface as long as one of the phases forming a buried interface is transparent in the visible region.²⁷ Moreover, we extended ESFG spectroscopy to femtosecond time-resolved measurements, which enabled us to examine the ultrafast dynamics of the solute molecules at the interface through the temporal change of time-resolved interfacial electronic spectra. This early trial for femtosecond time-resolved measurements was done by Kentaro Sekiguchi and Pratik Sen.^28,29

4. Jump: heterodyne-detected ESFG and its application

Most optical spectroscopies are explained on the basis of the following Tayler expansion of the polarization (P) that is induced in the material by the electric field (E) of light:

where P⁽ⁿ⁾ and χ⁽ⁿ⁾ are the n^th-order polarization and n^th-order susceptibility of the materials, respectively. The sum-frequency generation (SFG) is the second-order optical process that provides information about χ⁽²⁾, which arises only from the region in which inversion symmetry is broken under dipole approximation, such as interfaces. In the ESFG spectroscopy described in the previous section, we measure the intensity of the ESFG signal emitted by P⁽²⁾, which is proportional to |P⁽²⁾|². Therefore, the obtained ESFG spectrum is the |χ⁽²⁾|² spectrum of the interfacial molecules, but not of the χ⁽²⁾ spectrum itself. In absorption spectroscopy, in contrast, we measure the loss of the intensity of the incident light after passing through the sample, which corresponds to the energy transferred from the light to the molecules through the light–matter interaction. This quantity corresponds to the imaginary part of χ (Imχ) and hence the linear absorption spectrum represents the Imχ⁽¹⁾ spectrum. Consequently, the ESFG spectrum cannot directly be compared to the absorption spectrum, even though it is possible to compare their intensity maximum when only a single electronic transition contributes to the observed band in the spectrum, as in the case of coumarin dyes. Therefore, in general, a careful fitting analysis is necessary to derive proper molecular information from the |χ⁽²⁾|² spectra because the interference of different spectral components heavily affects the spectrum. This is a common, serious problem of the spectra obtained using the intensity measurements of the nonlinear optical signal emitted in a new direction (i.e. homodyne detection). I knew this problem well through our previous studies of heterodyne detection in transient-grating-type impulsive Raman spectroscopy³⁰ and noticed that it was becoming problematic in interfacial nonlinear spectroscopy while using the ESFG spectroscopy that we had developed.

In the field of VSFG, Shen and coworkers reported an elaborate way to solve this problem by developing scanning phase-sensitive VSFG spectroscopy, in which they determined the phase of the VSFG signal light at each frequency to obtain the imaginary part (Imχ⁽²⁾) and the real part (Reχ⁽²⁾) of the vibrationally resonant χ⁽²⁾ spectra.^31,32 Prof. Shen visited Japan in December 2006 and gave an excellent talk about their pioneering scanning phase-sensitive VSFG at a symposium held at the IMS in Okazaki. We attended the symposium and were very impressed by his talk. In particular, it ignited Shoichi, driving our lab to the heterodyne detection of SFG spectroscopy. I heard that Shoichi and Sobhan discussed Shen's lecture on the way back from Okazaki to Wako, and Sobhan strongly encouraged Shoichi to seriously pick up this problem.

After a while, Shoichi realized broadband multiplex heterodyne detection of ESFG (HD-ESFG) at RIKEN by constructing a setup as shown in Fig. 3a, which enabled measuring the electronically resonant Imχ⁽²⁾ spectrum (as well as Reχ⁽²⁾ spectrum) of interfacial molecules.³³ In this setup, the optical configuration of HD-ESFG is identical to the homodyne-detected ESFG setup up until the liquid interface irradiated by the ω₁ and ω₂ pulses. The difference is the additional configuration that generates the second sum-frequency signal from the GaAs surface, which is used as the local oscillator (LO) for the heterodyne detection. The ESFG signal from the liquid interface and the LO from the GaAs propagate collinearly and enter a polychromator, in which the sum of the ESFG signal and the LO are spectrally dispersed, interfere, and are detected by using a charge-coupled device (CCD). The typical raw data obtained are shown in Fig. 3b, in which the fringes arising from the interference between the ESFG signal and the LO are clearly seen. The phase and the amplitude of the fringe provide information about the phase and amplitude of the electric field of the ESFG signal, respectively. The Fourier analysis of this fringe finally provides the electronically resonant Imχ⁽²⁾ and Reχ⁽²⁾ of interfacial molecules, after calibrating the phase and amplitude using data obtained from a standard sample such as z-cut quartz.³³

One of the advantages of HD-ESFG spectroscopy is that it can determine the sign of the electronic band in the Imχ⁽²⁾ spectra, which contains direct information about the “up” or “down” alignment of interfacial molecules. Figure 3c shows the Imχ⁽²⁾ spectra of N,N-diethyl-p-nitroaniline (DEPNA) and p-nitroaniline (PNA) at the air/water interface, which were measured by using HD-ESFG.³⁴ These bands are due to the electronic transition to the lowest one- and two-photon allowed singlet excited state of each molecule but they appear with opposite signs, indicating that their absolute orientation at the air/water interface is opposite. At the air/water interface, it is readily imagined that the hydrophobic part of a solute molecule points upward to the air and the hydrophilic part downward to bulk water. In DEPNA, the diethylamino group is nonpolar and hydrophobic whereas the nitro group is polar and hydrophilic, so it is expected that DEPNA takes “nitro-group down” alignment at the air/water interface. In PNA, on the other hand, the nitro and amino groups are both polar and hydrophilic but the amino group is more hydrophilic, so the molecule is expected to take the opposite “amino-group down” alignment. In fact, the opposite signs of the χ⁽²⁾ spectra, as well as our complementary molecular dynamics (MD) simulation, indicate that this is the case. The 2 solute molecules at the air/water interface are “half solvated” in a significantly different manner, as governed by the balance of the hydrophilicity/hydrophobicity of the functional groups at the para positions.

The most important advantage of HD-ESFG is that it provides interfacial electronic spectra that can be directly compared with the linear absorption spectra in the bulk. By taking full advantage of this merit, we discussed pH at several water interfaces by measuring the electronic spectra of a pH indicator molecule at the interface by using HD-VSFG.^35–38 Figure 4 shows the electronic Imχ⁽²⁾ and Reχ⁽²⁾ spectra of a surface-active pH indicator dye, 4-heptadecyl-7-hydroxycoumarin, at the air/water interface.³⁶ This indicator dye is in the equilibrium between the neutral and anionic forms that exhibit significantly different electronic transitions. As shown in the figure, the Imχ⁽²⁾ and Reχ⁽²⁾ spectra of the indicator at the air/water interface drastically change with the change in the pH in the bulk water, reflecting the change in the acid–base equilibrium at the air/water interface, analogously to “litmus paper.” In fact, in the Imχ⁽²⁾ spectra that correspond to absorption spectra, the contribution from the acid (HA) is predominant in the wavelength range of shorter than 355 nm at the bulk pH lower than 11, whereas a peak due to the conjugate base (A^–) at 365 nm becomes significant at the bulk pH higher than 11. I emphasize that the Imχ⁽²⁾ and Reχ⁽²⁾ spectra clearly exhibit isosbestic points at 348 and 382 nm, respectively, which can never be observed in the |χ⁽²⁾|² spectra. By using the analysis, taking account of the finite concentration of the pH indicator as well as the change in its pKa value at the water interface, we estimated the pH of the air/water interface to be lower than the bulk pH by 1.7 based on the observed pH change of the Imχ⁽²⁾ and Reχ⁽²⁾ spectra.³⁶ This study demonstrated that we can perform spectrometry, e.g. pH spectrometry, at the interface in a similar fashion to the bulk using HD-ESFG. I stress that such spectrometry cannot be performed using traditional homodyne detection because the spectra cannot simply be separated into constituent spectral components.

Using the same strategy, we also evaluated the pH at the surfactant/water interface³⁵ and lipid/water interfaces.^37,38 In particular, in a study on lipid/water interfaces,³⁸ we provided a unified view for predicting the local pH at biomembranes based on the pH values evaluated at the water interfaces with the positively charged, negatively charged, nonionic zwitterionic lipids monolayers.

5. Expansion: heterodyne-detected VSFG and its application

The success of developing the multiplex HD-ESFG method boosted our liquid interface study and pushed us to extend it to VSFG. As I wrote in Section 3, I had decided not to start our interface research by using VSFG because VSFG had been utilized extensively by many research groups all over the world. However, it was obvious that we could develop multiplex broadband HD-VSFG based on HD-ESFG by changing ω₂ from white-light continuum to the broadband infrared (IR) pulse generated from an optical parametric amplifier and that HD-VSFG could provide vibrational Imχ⁽²⁾ of interfacial molecules, solving all the intrinsic problems of the |χ⁽²⁾|² spectra obtainable using conventional VSFG based on homodyne detection. Furthermore, it was readily expected that multiplex HD-VSFG would provide Imχ⁽²⁾ spectra with a much higher S/N and a much shorter measurement time than scanning phase-sensitive VSFG, which Shen and coworker realized. Therefore, there seemed no more reason not to start the interface research on VSFG by introducing the heterodyne-detection scheme that we developed for HD-ESFG.

5.1 Development of HD-VSFG and solving long-lasting questions

The development of HD-VSFG was realized by Satoshi Nihonyanagi and Shoichi³⁹ soon after the development of HD-ESFG by Shoichi. Satoshi had performed VSFG studies for his PhD thesis research as well as his postdoctoral research, and hence he had plenty of knowledge of the technique and research interests of the VSFG community. He joined RIKEN, planning to do research using the χ⁽⁴⁾ spectroscopy that we had developed, but he changed the plan and became the engine for our steady-state HD-VSFG studies.

Figure 5 shows the χ⁽²⁾ and |χ⁽²⁾|² spectra of the air/water interface, air/negatively charged surfactant (SDS) monolayer/water interface, and air/positively charged surfactant (CTAB) monolayer/water interface in the OH stretch region, which we reported in our first HD-VSFG paper.³⁹ In the Imχ⁽²⁾ spectrum of the air/water interface (Fig. 5a), the 2 OH stretch bands can be seen. The sharp positive OH band at 3,700 cm⁻¹, denoted by OH(III), is attributed to the OH stretch vibration of the dangling OH (free OH) that sticks out into the air side at the air/water interface. The broad negative OH band of around ∼3,450 cm⁻¹, denoted by OH(II), is due to the vibration of the hydrogen-bonded OH (H-bonded OH) of interfacial water. Importantly, the former appears with a positive sign whereas the latter has a negative sign, directly reflecting the absolution orientation of the OH moiety: free OH has “H-up” orientation whereas H-bonded OH has “H-down” orientation on average at the air/water interface, being very consistent with our intuition. (The third band, denoted by OH(I), will be discussed in Section 5.3). On the other hand, at the air/surfactant monolayer/water interfaces, an H-bonded OH stretch band appears in the Imχ⁽²⁾ spectrum with a positive sign when the surfactant has a negatively charged head group (SDS, Fig. 5b). This is because the H atom of water is slightly positive and has a tendency to be located closer to the negatively charged interface, making the H-up orientation energetically more favorable. In contrast, an H-bonded OH stretch band appears with a negative sign in the Imχ⁽²⁾ spectrum when the surfactant has the positively charged head group (CTAB, Fig. 5c) because the H-down orientation is more energetically favorable at the positively charged water interface. These Imχ⁽²⁾ spectra proved the flip-flop model of interfacial water, in that the orientation of interfacial water changes depending on the sign of the charge at the interface.⁴⁰ The sign information of the band is obtained from the phase information of the nonlinear optical signal, as in the case of HD-ESFG, and it provides clear information about the absolute orientation of the relevant OH moiety. This highlights an advantage of HD-VSFG over conventional VSFG with homodyne detection.

Although our first HD-VSFG paper was not appreciated by the reviewers of general chemistry journals, we were convinced that HD-VSFG was a new, very powerful tool for investigating the fundamental problems of interfaces because the obtained Imχ⁽²⁾ spectra are interface-selective vibrational spectra, free from any spectral distortion arising from the square nature of the |χ⁽²⁾|² spectra of conventional VSFG. In fact, vibrational Imχ⁽²⁾ spectra can be straightforwardly interpreted and can be compared directly with IR and Raman spectra, which correspond to Imχ⁽¹⁾ and Imχ⁽³⁾ spectra, respectively. HD-VSFG soon provided key data for solving long-lasting questions related to interfacial water. For instance, we compared H-bonded OH stretch bands in the Imχ⁽²⁾ spectra of interfacial water at the negatively charged SDS and positively charged CTAB monolayers with the IR spectra of bulk water, using isotopically diluted water (HOD-D₂O) for which the OH stretch band predominantly arises from the HOD specie that is free from the intra- and intermolecular vibrational couplings.⁴¹ The obtained Imχ⁽²⁾ spectra showed that the OH stretch bands of these charged water interfaces are essentially the same as the OH stretch band of the IR spectra of bulk HOD-D₂O. This result clearly denied the presence of the “ice-like” water structure at the charged water interface, although it had been claimed, based on the homodyne-detected VSFG spectra, that the electric field induced an ice-like water structure at the interface.^42,43

One of the highlights in this line of our studies is the proposal of the molecular-level mechanism of the Hofmeister series of the counterions effect, which has been a century-long mystery in the community. The Hofmeister series was originally proposed to describe the order of the ability of ions to precipitate proteins out of solutions, named after Hofmeister, who first noticed this phenomenon in the late nineteenth century. For example, the ions on the left side of the Hofmeister series in Fig. 6a stabilize the native structure of proteins and salt out proteins, whereas the ions on the right side denature and solubilize proteins.^44–47 This order of the ion effect has been found not only in protein precipitations, but also in a variety of macroscopic phenomena such as the surface tensions of electrolyte solutions,⁴⁸ phase transitions of simple surfactants,^49,50 etc. We chose a positively charged cetyltrimethylammonium bromide (CTAB) monolayer/electrolyte solution interface and a negatively charged sodium dodecyl sulfate (SDS) monolayer/electrolyte interface as the model systems, and measured the Imχ⁽²⁾ spectra of the interfacial water using HD-VSFG with the addition of various excess salts. The obtained spectra indicated that, at the positively charged CTAB interface, the amplitude of the OH stretch band of the interfacial water decreases with the addition of counter anions in the order of the Hofmeister series (Fig. 6b). Because the OH stretch band amplitude reflects the net orientation of the interfacial water that is determined by the surface charge, this result indicates that the adsorption of the halide anion onto the positively charged interface determines the Hofmeister order of negative ions, as previously proposed by Cremer and coworkers.⁵¹ At the negatively charged interfaces, on the other hand, the counter cations did not drastically change the OH stretch band amplitude, but systematically altered its peak frequency (Fig. 6(c)), indicating that they affect the hydrogen-bond strength of the interfacial water in a way that correlates with the Hofmeister order of the cation. This study showed that the order of the halide anion effects at the positively charged interface can be explained by the propensity of the ion for adsorption whereas the order of the cation effects at the negatively charged interfaces arises from the difference in the effect on the hydrogen-bond strength of interfacial water. We proposed that the hydrogen-bond strength of the water at the interfaces is important in addition to the adsorption propensity of the ions, and that different mechanisms compositely play roles in order to realize the anionic and cationic Hofmeister series.⁵²

5.2 Interfacial water at the model biological membrane (lipid monolayer/water interfaces)

An important application of HD-VSFG is the elucidation of the water structure at the biological membrane interface. Jahar Alam Mondal initiated this direction of the HD-VSFG studies in our laboratory. Taking a leave of 2 yr from his home institute in India, Alam joined our laboratory with full energy and his own idea on time-resolved absorption studies. Soon, however, he noticed the importance of HD-VSFG and made up his mind to devote himself to the study of the model membrane by using it. Biological membranes are semipermeable barriers that separate a cell or cellular organelles (nucleus, mitochondria, etc.) from the surroundings, and this separation is essential for maintaining the environment inside the organelles and cells for their proper functioning. The function of membranes and membrane proteins is crucially dependent on the aqueous environment at the inner and outer surfaces of the membranes. This is because, for example, the membrane–water interaction affects the membrane electrostatics that regulates vital processes, such as signal transduction, as well as the transport of ions, drugs, and biomolecules across the membrane. Therefore, elucidation of the molecular-level structure and physicochemical properties of water at the membrane/water interface is important. Membranes are complex assemblies of lipids, proteins, carbohydrates, and cholesterols, so direct probing of the real membrane/water interface is difficult. Thus, we studied the interfacial water structure of model systems, i.e. lipid monolayers on the water surface, by using HD-VSFG.

Figure 7 shows the Imχ⁽²⁾ spectra of several lipid monolayer/water interfaces that we measured, with the chemical structure of the lipids. We used isotopically diluted water (HOD-D₂O) to make spectra simple and interpretation straightforward by suppressing intra/intermolecular vibrational couplings. As shown in Fig. 7b, when the lipid is negatively charged (anionic lipid) or positively charged (cationic lipid), the OH stretch band of the interfacial water appears with a positive or negative sign, respectively.⁵³ This demonstrates that a flip-flop of the interfacial water occurs, depending on the sign of the charge at the water interface, showing that the net orientation of interfacial water is governed by the net charge on the lipid head group, as in the case of simpler charged surfactant/water interfaces (Fig. 5).³⁹ An intriguing finding for the lipid/water interface was a peculiar spectral feature of the OH stretch band in the Imχ⁽²⁾ spectrum of the neutral zwitterionic phosphatidylcholine/water interface: although this lipid interface is net neutral, it exhibits a positive OH stretch band with a dent feature at around 3,500 cm⁻¹ (red in Fig. 7c). To interpret this peculiar spectrum of the zwitterionic lipid (DPPC)/water interface, we also measured the Imχ⁽²⁾ spectra of a zwitterionic surfactant (DDAO)/water interface as well as mixed lipid/water interfaces. The Imχ⁽²⁾ spectrum of the DDAO/water interface clearly shows both positive and negative bands in the OH stretch region (blue in Fig. 7c), revealing that multiple water structures having opposite orientations exist at the interface. For the mixed lipid/water interfaces, actually, while varying the fraction of the anionic and cationic lipids, we observed a gradual change in the Imχ⁽²⁾ spectra, in which spectral features that are similar to the anionic, zwitterionic, and cationic lipid/water interfaces appear successively. These observations indicated that, when the positive and negative charges coexist at the interface, H-down-oriented water structure and H-up-oriented water structure appear in the vicinity of the respective charged sites of the head group of the lipid. Based on these observations, we concluded that the characteristic Imχ⁽²⁾ spectrum of the zwitterionic lipid/water interface arises from 3 different types of water existing at the interface: (1) the water associated with the negatively charged phosphate, which is strongly hydrogen-bonded and has a net H-up orientation; (2) the water around the positively charged choline, which forms weaker hydrogen bonds and has a net H-down orientation; and (3) the water that is weakly interacting with the hydrophobic region of the lipid, which has a net H-up orientation.⁵⁴

5.3 Spectrum of the air/water interface: a big mistake and correction

Our study using HD-VSFG was going very well, including its extension to time-resolved measurements, which I describe in the next section. However, we suddenly had a very bitter lesson. In January 2015, I received a phone call from Shoichi Yamaguchi, who had left my laboratory in 2014 and started his own group as a full professor at Saitama University. In his new laboratory, he started developing a novel scanning HD-VSFG technique to measure Imχ⁽²⁾ and Reχ⁽²⁾ spectra at the interface using a picosecond laser system (“Saitama” method), so he measured the Imχ⁽²⁾ spectrum at the air/water interface again to check the reliability of his new method. In the phone call, he said that something was wrong. He told me that the low-frequency positive spectral feature in the Imχ⁽²⁾ spectrum at the air/water spectrum, denoted by OH(I) in Fig. 5a, disappeared when he used the Imχ⁽²⁾ spectrum at the air/D₂O interface as the reference for the phase and amplitude calibration of the HD-VSFG signal. Until this time, we had used the quartz spectrum as the reference for calibration without any doubt, but the air/D₂O spectrum could also be used as a reference for the OH stretch region because it did not show any vibrational resonance in the relevant frequency region. In fact, the air/D₂O spectrum is an even better reference for the measurement of the air/water interface because the optical alignment can be kept exactly the same for the measurements of both the sample (the air/H₂O interface) and the reference (the air/D₂O interface). The information provided by Shoichi was totally unexpected and very astonishing because all the Imχ⁽²⁾ spectra of the air/water interface reported by that time, including the first Imχ⁽²⁾ spectrum reported by the Shen group⁵⁵ as well as that in our first HD-VSFG paper (Fig. 5a),³⁹ had shown a positive spectral feature in the region of <3,200 cm⁻¹,^56–59 and even several MD simulations reproduced it.^60,61 Furthermore, this positive spectral feature in the low-frequency region of the OH stretch region was intensively discussed, relating to the existence of the “ice-like” water structure at the air/water interface.⁶²

I was shocked by Shoichi's phone call. Although I had sometimes asked my coworkers in the laboratory about the reproducibility of the positive spectral feature and heard from everybody that the positive feature was real, I could not overlook his phone call because I fully trusted Shoichi's straightforwardness and honesty in science. So, I immediately stopped all the other projects of our interface research that were running in the laboratory and requested all members in the interface group to check the reproducibility of the positive spectral feature in the low-frequency OH stretch region of the air/water interface.

Figure 8 (red) shows the Imχ⁽²⁾ spectrum of the air/water interface obtained after very careful, repeated measurements. As can be clearly seen, there is no positive spectral feature in the region of <3,100 cm⁻¹. Based on a thorough examination, we concluded that the positive feature in the low-frequency OH stretch region that had been reported before was an artifact that arose from the inaccurate phase calibration to obtain the Imχ⁽²⁾ spectrum, i.e. an error in the phase correction or an artifact originating from dust in the quartz used for the phase reference. We rushed to publish to correct the big mistake that we had made and provided the reliable Imχ⁽²⁾ spectrum at the air/water interface with Shoichi,⁶³ who also independently published the correct Imχ⁽²⁾ spectrum with the development of his new scanning HD-VSFG technique.⁶⁴ Our correction of the Imχ⁽²⁾ spectrum at the air/water interface surprised the community and ignited some debates initially,^65–68 but the absence of the vibrationally resonant positive spectral features in the low-frequency OH stretch region is now established. The correct Imχ⁽²⁾ spectrum at the air/water interface also negates the existence of “ice-like” water structure at the air/water interface.

I had thought that we were doing research very carefully and understood the possible errors included in our data well. However, this mistake about the Imχ⁽²⁾ spectrum of the air/water interface taught me that it was not enough. On reflection, there had been chances for me to notice this error earlier. For example, our first Imχ⁽²⁾ spectrum of the air/water interface (Fig. 5a) also exhibits a positive offset in the high-frequency edge beyond the free OH band at 3,700 cm⁻¹. Because the constant offset is a typical feature arising from the contamination of the Reχ⁽²⁾ component to the Imχ⁽²⁾ spectrum due to the error in phase calibration, I would have been able to realize this problem if I had only asked myself a simple question: “How high a frequency does the tail of the free OH band extend to?” or “How low a frequency does the positive spectral feature extend to?” This was a bitter, but very instructive, lesson that I learned during our research using steady-state HD-VSFG.

6. Challenge: extension of heterodyne-detected VSFG to time-resolved measurements

The studies described in the previous section demonstrated marked advantages of HD-VSFG over conventional VSFG performed with homodyne detection. The sign information of Imχ⁽²⁾ providing information about the absolute orientation of interfacial molecules is one of them, which we stressed in our early HD-VSFG studies.^39,53 However, I consider that the most important advantage of HD-VSFG is that it affords the additivity of spectral components because the measured signal is linearized to χ⁽²⁾. In other words, when the observed spectrum $Imχtotal(2)$ consists of multiple spectral components $Imχi(2)$⁠, the total spectrum can be represented as a simple sum of the constituent components:

This advantage of HD-VSFG becomes even more crucial for time-resolved measurements, in which we observe the spectral change induced by pump pulse irradiation. In fact, time-resolved VSFG with homodyne detection provides the pump-induced change in |χ⁽²⁾|², namely:

where χ⁽²⁾_steady and Δχ⁽²⁾(t) are steady-state χ⁽²⁾ and pump-induced change of χ⁽²⁾ at the delay time of t, respectively. As seen in Equation (3), the physical meaning of the observed time-resolved signal is complicated, and obviously it is very difficult to interpret the time-resolved VSFG spectra measured using homodyne detection. In contrast, time-resolved spectra obtained by using HD-VSFG are just the photoinduced change in χ⁽²⁾:

Thus, time-resolved spectra obtained using time-resolved HD-VSFG measurements can be interpreted straightforwardly, in the same way as we interpret time-resolved spectra measured using time-resolved IR/Raman spectroscopy in solution. With this thinking, we extended HD-VSFG to time-resolved measurements to explore ultrafast dynamics at the liquid interface, which looked very intriguing but obviously unexplored.

6.1 Ultrafast vibrational dynamics at water interfaces: IR-excited time-resolved HD-VSFG and 2D HD-VSFG

When I was a young research associate at KAST, I studied the photoisomerization of retinal molecules in solution using picosecond time-resolved coherent anti-Stokes Raman spectroscopy (CARS) that provides Δ|χ⁽³⁾|².⁶⁹ Hence, I knew well the difficulty in interpreting time-resolved Δ|χ⁽³⁾|² spectra measured using homodyne detection even though the spectra themselves could be measured using high S/N. In this sense, the advantage of HD-VSFG for time-resolved measurements was obvious to me when steady-state HD-VSFG was realized at RIKEN. Moreover, the extension of femtosecond laser-based HD-VSFG to femtosecond time-resolved measurements did not technically look difficult because it could be done only by adding one more femtosecond pulse for photoexcitation. So, I immediately started encouraging my coworkers not to stay with steady-state measurements, but to proceed swiftly to time-resolved measurements. Actually, at around that time in 2009, Satoshi Nihonyanagi and I had a chance to attend a conference held on Ischia island in Italy together; I enthusiastically explained to him the importance of heterodyning in the time-resolved measurements during the trip and persuaded him to try time-resolved experiments. A while later, Satoshi succeeded in the first demonstration of time-resolved HD-VSFG (TR-HD-VSFG) measurement and showed its high potential in revealing ultrafast vibrational dynamics at water interfaces.⁷⁰ In this way, we also started studying ultrafast vibrational dynamics at interfaces using IR-excited TR-VSFG and 2D HD-VSFG soon after the realization of steady-state HD-VSFG.

The main contributors to the extension of HD-VSFG to IR-excited time-resolved measurements were Prashant C. Singh,^71–74 Kenichi Inoue,^75–79 Ahmed Mohammed,^80,81 and Woongmo Sung,⁸² with the guidance of Satoshi and/or Shoichi. In particular, Prashant's enthusiasm opened the door to 2D HD-VSFG. He was an energetic postdoc who said to me in his job interview: “I want to do difficult things.” He devoted himself to realizing 2D spectroscopy using HD-VSFG first in our laboratory.

The vibrational dynamics of water, including the fluctuation in its hydrogen bonding, is one of the major research subjects in ultrafast spectroscopy for the bulk. Thus, we naturally set the target of IR-excited TR-HD-VSFG at the vibrational dynamics of interfacial water. Although several pioneering attempts have been made by using time-resolved VSFG with homodyne detection,^83–88 the studies on ultrafast vibrational dynamics of interfacial water at the bulk level were not realized. It became possible only with the development of IR-excited TR-HD-VSFG.

Figure 9 shows typical interface-selective time-resolved vibrational spectra obtained using IR-excited TR-HD-VSFG.⁷⁶ These ΔImχ⁽²⁾ spectra were obtained from the air/water interface with vibrational excitation of low-frequency (3,300 cm⁻¹) and high-frequency (3,500 cm⁻¹) sides of the broad H-bonded OH stretch band. Since the bandwidth of the femtosecond IR pump pulse (∼120 cm⁻¹) is much narrower than the bandwidth of the H-bonded OH stretch band (top traces in Fig. 9), the IR pump pulse can selectively excite a part of the broad OH stretch band. As shown in these time-resolved spectra, narrow bleaches are observed at different frequencies immediately after excitation (0.0 ps), reflecting the difference in the pump frequency (3,300 and 3,500 cm⁻¹). (Note that the bleach of the H-bonded OH stretch band, i.e. the decrease in the amplitude of the band, appears with a positive sign in time-resolved ΔImχ⁽²⁾ spectra because this band has a negative sign in the steady-state Imχ⁽²⁾ spectra.) In other words, the ΔImχ⁽²⁾ spectra exhibit spectral holes (hole burning), demonstrating that the broad H-bonded OH stretch band at the air/water interface is inhomogeneously broadened. These time-resolved ΔImχ⁽²⁾ spectra measured using the 2 different IR excitations became very similar to each other at 0.5 ps and were indistinguishable at 1.0 ps. This implies that, although there exists spectral inhomogeneity in the H-bonded OH stretch, efficient spectral diffusion due to fast hydrogen-bond fluctuation occurs and the memory of the inhomogeneity is washed out on the sub-picosecond timescale regardless of the initial frequency in the H-bonded OH stretch vibration. Around that time, it had been claimed that there was energetically isolated H-bonded OH at around 3,500 cm⁻¹ that shows extremely slow spectral diffusion (>1.5 ps) at the air/water interface.⁸⁹ The time-resolved ΔImχ⁽²⁾ measured using TR-HD-VSFG spectroscopy unambiguously negated it and revealed the correct vibrational dynamics of the air/water interface. TR-HD-VSFG has proven its powerfulness for elucidating the vibrational dynamics at the liquid interface through a series of studies for the air/water interface.^{72,76,78,81,82}

For the study of ultrafast dynamics in the bulk, 2D IR is one of the most powerful advanced methods and has been utilized extensively.⁹⁰ In 2D IR, simply speaking, the temporal evolution of femtosecond time-resolved IR absorption spectra and its IR pump frequency dependence are measured, and the obtained spectral information is presented in the form of 2D spectra, revealing a variety of vibrational dynamics of the systems such as vibrational relaxation, energy/population transfer, inhomogeneity and spectral diffusion, chemical exchange, and so on. To thoroughly investigate vibrational dynamics at liquid interfaces using TR-HD-VSFG, we can also perform 2D spectroscopy by measuring time-resolved ΔImχ⁽²⁾ spectra by systematically changing the IR pump frequency. The obtained 2D HD-VSFG spectra for liquid interfaces are essentially equivalent to 2D IR spectra for the bulk, although the signal sign in 2D HD-VSFG changes with the change in the absolute orientation of the moiety that gives a relevant vibrational band.

The strength of 2D HD-VSFG was also highlighted by a series of studies on lipid/water interfaces. Figure 10 depicts 2D HD-VSFG spectra at the negatively charged 1,2-dipalmitoyl-sn-glycero-3-phosphorylglycerol (DPPG) monolayer/water (HOD-D₂O) interface and the positively charged 1,2-dipalmitoyl-3-trimethylammonium propane (DPTAP) monolayer/water (HOD-D₂O) interface in the H-bonded OH stretch region.⁷⁴ (Note that we used isotopic diluted water to suppress the complexity arising from intra- and intermolecular coupling.) In each 2D spectrum, the time-resolved ΔImχ⁽²⁾ spectrum at a delay time is plotted along the horizontal axis (ω₂) and the IR pump frequency used for the measurement of this ΔImχ⁽²⁾ spectrum is indicated in the vertical axis (ω_pump). DPPG has a negatively charged head group that contains a phosphate group (–PO₄⁻–), so the interface is negatively charged and the OH stretch band of interfacial water appears with a positive sign (H-up orientation). Thus, the bleach of the OH stretch band appears with a negative sign, giving rise to the negative “bleach lobe” in the 2D spectra at the DPPG monolayer/water interface (Fig. 10a). As shown in the figure, the bleach lobe at the delay time of 0.0 ps is clearly tilted, reflecting that the frequency of the spectral hole (plotted along the horizontal ω₂ axis) changes with the change in the IR pump frequency (plotted along the vertical ω_pump axis). This is a clear manifestation of the significant inhomogeneous broadening of the OH stretch band of water at the DPPG monolayer/water interface. As the delay time increases, the memory of the initial pump frequency is gradually lost due to the hydrogen-bond fluctuation (spectral diffusion), which changes the tilted bleach lobe to round on a sub-picosecond timescale. For comparison, Fig. 10b shows 2D HD-VSFG spectra at the DPTAP monolayer/water interface, which is positively charged due to the choline group (–N(CH₃)₃⁺) in the head group. The bleach lobe in the 2D HD-VSFG spectra is almost vertical even at the delay time of 0.0 ps, indicating that spectral diffusion at the DPTAP/water interface is so fast that it is almost finished within the time resolution of the measurements (∼200 fs). The obtained 2D HD-VSFG spectra suggest that hydrogen-bond fluctuation of interfacial water at the DPPG/water interface is much slower than that at the DPTAP/water interface, and the latter is very similar to that of bulk water. This distinct difference in the hydrogen-bond fluctuation at the oppositely charged lipid/water interfaces was rationalized by the difference in the interaction between interfacial water and the lipid head group: the phosphate group of DPPG forms firm hydrogen bonds with interfacial water, and glycerol OHs at the terminal end can also provide additional hydrogen-bonding sites, which largely limits the fast fluctuation of interfacial water at the DPPG/water interface. On the other hand, the choline group of DPTAP only forms weak hydrogen bonds with water, so the OH of the interfacial water can fluctuate much more freely at the DPTAP/water interface. Although the steady-state Imχ⁽²⁾ spectra can only indicate the similar averaged hydrogen-bond strength of interfacial water at the DPPG/water and the DPTAP/water interfaces, the 2D spectra clearly showed that its hydrogen-bond dynamics is very different, as illustrated in the bottom sketches in Fig. 11a and c. This demonstrates the powerfulness of 2D spectroscopy to reveal the essential properties of interfacial water from a dynamic viewpoint.

2D HD-VSFG spectroscopy on interfacial water at the zwitterionic phosphatidylcholine (DPPC) interface, which has both anionic phosphate and cationic choline in the head group, provided further insights into the lipid/water interface.⁷⁹ Figure 11b depicts the 2D HD-VSFG spectrum at the DPPC monolayer/water (HOD-D₂O) interface at the delay time of 0.0 ps, i.e. immediately after the IR pump. In addition to the large negative bleach lobe (blue) of the OH stretch band centered at around 3,300 cm⁻¹, an additional small positive lobe (red) is observed in the high ω₂ frequency region (∼3,500 cm⁻¹). The appearance of this additional minor positive lobe in the high ω₂ frequency region is clear evidence that there exist 2 OH stretch bands with opposite signs in the OH stretch region at the zwitterionic DPPC/water interface, although the OH stretch band in the steady-state Imχ⁽²⁾ spectrum is overall positive (Fig. 7). Indeed, this small positive lobe is attributable to the bleach lobe of a negative OH stretch band of H-down water molecules around the cationic choline site in the head group. The 2D HD-VSFG spectrum unambiguously showed that multiple water species having different orientations are present at the zwitterionic lipid DPPC/water interface, as we concluded based on a steady-state HD-VSFG study (Section 5.2).⁵⁴ It is noteworthy that there were some debates before this 2D HD-VSFG experiment: the conclusion of our steady-state HD-VSFG study was supported by several MD simulations including ours,^91,92 whereas several MD studies denied it or did not recognize the H-down water around the choline site.^93–95 Another important finding for the 2D spectrum of the DPPC/water interface is that both of the 2 bleach lobes are almost vertically elongated even at 0.0 ps, indicating ultrafast loss of the memory of the initial pump frequency of the OH stretch vibration. In particular, the tilt of the main negative bleach lobe centered at 3,300 cm⁻¹ in the 2D spectrum of the DPPC/water interface (Fig. 11b) is very small compared with the tilt of the negative bleach lobe in the 2D spectrum of the DPPG/water interface (Fig. 11a), although both negative lobes arise from the OH stretch of the water molecules that hydrate the phosphate head group with H-up orientation. This implies that the ultrafast fluctuation of the interfacial water hydrogen-bonded to the phosphate head group is not efficiently suppressed at the zwitterionic lipid interface because of the disturbance of the rapidly fluctuating OH of water around the neighboring choline group.⁷⁹

While carrying out the studies on ultrafast vibrational dynamics at aqueous interfaces, we became convinced that IR-excited TR-HD-VSFG and 2D HD-VSFG are very powerful in revealing ultrafast vibrational dynamics of interfacial water. Actually, we found that the “hidden isolated OH” exists at a negatively charged hydrophobic aqueous interface but not at a corresponding positive hydrophobic aqueous interface, revealing the importance of the water orientation in the interaction at the interface.⁸⁰ Furthermore, we succeeded in obtaining a unified view of the vibrational relaxation of the OH stretch vibration at the air/water interface very recently, including that of the free OH that is characteristic of the air/water interface.⁸²

6.2 Ultrafast reaction dynamics at liquid interfaces: UV-excited time-resolved HD-VSFG

Along with the study of vibrational dynamics at the interface using IR pump TR-HD-VSFG and 2D HD-VSFG, we also started attempts to investigate reaction dynamics at the interfaces using TR-HD-VSFG.

It is widely known that liquid interfaces provide unique environments for important reactions in organic/inorganic chemistry, electrochemistry, environmental science, biological science, and others. In particular, chemical reactions at the water interface are essential in environmental chemistry because many key reactions occur at the aerosol surfaces as well as the sea surface. Several experimental and theoretical studies suggested that chemical reactions at the water interfaces are significantly different from those in the bulk water. For instance, it was reported that the rates of some chemical reactions drastically increase in an emulsion, which is called “on-water catalysis,” and the essence of this on-water catalysis has been considered to be the acceleration of the reaction at the water interface.^96–98 Although such indirect information suggesting the uniqueness of reactions at the water interface was reported, it is very challenging to directly investigate chemical reactions at the water interface and hence molecular-level understanding of interfacial reactions is very scant.

I first noticed this intriguing subject when I attended the Faraday Discussion on Frontier in Physical Organic Chemistry in 2009,²⁹ where Prof. Marcus gave a talk about his theoretical study of on-water catalysis.⁹⁹ I was very impressed by his talk and asked a question about the possibility of the study using interface nonlinear spectroscopy (which was recorded and published).¹⁰⁰ Since then, I wished to take on the challenge of directly observing chemical reactions at the interface in a time-resolved manner. Then, finally, tracking ultrafast chemical reactions and short-lived transients at the water interface (or water surface) was attempted and realized by Korenobu Matsuzaki and Ryoji Kusaka.^101,102

To track chemical reactions proceeding at the water surface, we developed UV-excited TR-HD-VSFG spectroscopy that utilizes ultraviolet pump pulses to start photochemical processes. In principle, this UV-excited TR-HD-VSFG experiment can be realized by replacing IR pump pulses in IR-excited TR-HD-VSFG with the UV pump pulses. In practice, however, we needed to solve the problem of the photoproducts accumulation and suppress the degradation of the reactant molecules by making a new flow cell system that continuously flows the aqueous solution while maintaining its surface height with micrometer accuracy to realize a high phase reliability in the TR-HD-VSFG measurements. Using this new setup, we first tried to observe hydrated electrons at the water surface. The hydrated electron is the simplest ionic species and a very highly reactive transient that plays essential roles in photochemistry and radiochemistry. Because of its importance, the hydrated electron has been extensively studied in bulk water using time-resolved absorption spectroscopy by monitoring its famous transient absorption peak at ∼720 nm.¹⁰³ Although the study of vibrational spectroscopy is rare, we previously measured time-resolved spontaneous Raman spectra under the resonance condition with transient absorption of the hydrated electrons and discovered that the vibrations of the water molecules that hydrate electrons exhibit strong Raman bands of around 2000.^104–106 Because we had failed to observe the characteristic electronic transition of hydrated electrons at the interface using time-resolved (homodyne) ESFG spectroscopy before, we bet on the possibility of detecting the OH stretch band of interfacial water hydrating to the electrons using UV-excited TR-HD-VSFG by ejecting electrons onto the surface of the water with the photoionization of water or solute molecules by UV pump pulses.¹⁰¹

Figure 12a shows time-resolved ΔImχ⁽²⁾ spectra in the OH stretch region at the air/water interface observed by introducing UV pump pulses that photoionize water molecules and eject electrons onto the water surface. A new, strong transient band appears with a positive sign immediately after the UV excitation. This transient band is assignable to the OH stretch of the interfacial water that directly interacts with electrons at the air/water interface. It appears with a positive sign, indicating that the relevant OH bond has a net H-up orientation, implying that the water molecule is associated with an electron from the bottom (from the bulk water side). This transient band disappears on a picosecond timescale, indicating that hydrated electrons are not stable at the water surface and escape into the polar bulk water to gain more stabilization energy through full hydration. Figure 12b compares the vibrational spectrum of “hydrated electrons” at the air/water interface with those of a water anion cluster, (H₂O)₆^– in the gas phase (top)¹⁰⁷ and hydrated electrons in the bulk water (bottom).¹⁰⁵ The water anion clusters have been intensively studied as a model for hydrated electrons¹⁰⁸ and it is considered that the electron of small clusters is exposed to the vacuum. In (H₂O)₆^–, in particular, the electron is hydrated by one water molecule at the cluster surface with 2 OH groups pointing toward the electron, so 2 OH stretch vibrations mix and give rise to 2 OH bands at 3,265 and 3,376 cm⁻¹ by intramolecular vibrational coupling. Thus, the intrinsic frequency of the individual OH stretch modes is estimated by averaging the 2 frequencies to be ∼3,320 cm⁻¹. On the other hand, in bulk water, it is considered that an electron is surrounded by (about 4 to 6) water molecules and is fully hydrated, which shows a very broad band centering at around 3,170 cm⁻¹.¹⁰⁵ The comparison in Fig. 12b shows that the OH stretch frequency of the hydrated electrons at the air/water interface (3,260 cm⁻¹) is located in between the frequencies in the gas-phase cluster (3,320 cm⁻¹; exposed to the air) and liquid water (3,170 cm⁻¹; fully hydrated). This implies that electrons are “partially” hydrated at the air/water interface.¹⁰¹

The success of detecting the partially hydrated electron at the air/water interface led us to realize that we can detect transient species by observing the change in the hydration structure by monitoring the OH stretch region even if the direct observation of the transients’ vibrations themselves is difficult. Encouraged by this idea, we proceeded to the tracking of photochemical reactions at the water surface. We picked up the photochemical reaction of phenol, which generates phenoxy radical, proton, and electron in bulk water (Fig. 13a), and attempted to measure time-resolved ΔImχ⁽²⁾ spectra in the OH region after photoexcitation of the phenol molecule by using femtosecond UV pulses at 267 nm.¹⁰²

Figure 13b shows time-resolved ΔImχ⁽²⁾ spectra at the delay times from −0.7 to 300 ps after UV excitation on the surface of an aqueous solution of phenol. Strong transient signals are observed in the regions of 2,800 to 3,400 and 3,400 to 3,700 cm⁻¹ with negative and positive signs, respectively. The temporal spectral change clearly shows that the time-resolved ΔImχ⁽²⁾ spectra contain the signals of 3 transients, all of which appear within the time resolution of the present measurements (∼200 fs) and monotonically decay on different timescales. These 3 transients are attributed to (1) partially hydrated electron e^–_hyd (∼3440-cm⁻¹ positive band, decaying on a sub-picosecond timescale), (2) proton H⁺ (the major part of ∼3200-cm⁻¹ negative band, decaying on a ∼100 ps timescale), and (3) phenoxy radicals PhO• (the rest of the 3200-cm⁻¹ negative band and ∼3600-cm⁻¹ positive band, showing no changes within the delay time range of the measurement, i.e. 300 ps). As far as I know, this is the first direct experiment to have tracked chemical reactions at the liquid interface. What surprised us was that the phenol photoreaction dynamics at the air/water interface were drastically different from those in bulk water, although the reaction itself was the same: the observed time-resolved ΔImχ⁽²⁾ spectra revealed that the photoproducts of the reaction, i.e. hydrated electron e^–_hyd, proton H⁺, and phenoxy radicals PhO•, are generated within ∼0.1 ps at the air/water interface after photoexciting phenol with the 267-nm UV pump pulse, whereas time-resolved absorption spectroscopy had reported that the same 267-nm excitation generated the PhO• transient slowly on a nanosecond timescale in the bulk water.¹⁰⁹ This means that the photochemical reaction of phenol becomes more than 10⁴ times faster than that in bulk water even though the initial photoexcitation photon energy is the same.

Because the result of the first experiment of tracking reactions on the water surface was very surprising, I wanted to rationalize this observation by using computation. The mechanism and dynamics of photochemical reactions are governed by the relevant potential energy surfaces and, in the case of the photochemical reaction of phenol, it is known that the conical intersection (CI) of S₁ (¹ππ*) and S₂ (¹πσ*) determines the energy barrier for the initial excited state to undergo a reaction.^110–112 Professor Ishiyama at University of Toyama and Prof. Morita at Tohoku University agreed to theoretically investigate this problem and performed quantum chemical calculations combined with MD simulation. Their computation quantitatively confirmed that the relative energy of S₁ (¹ππ*) and S₂ (¹πσ*) phenol substantially changes at the water surface, reflecting the unique “half-solvation” environment at the interface, and that the energy barrier for the reaction (i.e. ¹ππ*/¹πσ* CI energy relative to the Franck–Condon state) is substantially lowered compared with the bulk, making the photochemical reaction of phenol ultrafast at the air/water interface.¹¹³

The development of UV-excited TR-HD-VSFG spectroscopy enabled us to track a photochemical reaction at the air/water interface and to find that the photochemical reaction of phenol becomes ultrafast at the water interface. I consider that the difference in the reaction between the interface and the bulk solution is not specific to the photochemical reaction of phenol and that our finding has a much more general implication. The CI is an essential element that governs the mechanism and dynamics of chemical reactions. Because the CI is the point at which 2 electronic states cross at the potential energy surface, its location and energy shift as the relative energy of the 2 electronic states is changed. Because each electronic state usually has a different electronic character, the sensitivity of its energy to the environment is different. Consequently, the relative energy of the 2 electronic states is expected to be naturally different between the interface and the bulk, and hence the location of the CI as well as the barrier height that the CI generates is also expected to be different, which makes the reaction at the interface different from the bulk in general. In this sense, the success of tracking the ultrafast reaction using UV-excited TR-HD-VSFG spectroscopy may open a new research field, i.e. the study of chemical reactions and reaction dynamics at liquid interfaces.

7. Concluding remarks

After about 20 yr of struggle that started with naive, dreaming discussions at a dumpling restaurant in Shinjuku, I now feel that my initial dream has been realized, at least partially: bringing up the level of spectroscopic study at liquid interfaces to the bulk level. At the moment, however, the dream has been realized only for gas/liquid interfaces—in particular, simple air/aqueous and air/monolayer/aqueous interfaces. Without any doubt, there are an enormous number of important, unexplored problems for liquid interfaces that can be investigated using interface-selective nonlinear spectroscopy. Lastly, I wish to mention several possible future directions.

A straightforward direction is the study of the ultrafast dynamics at liquid interfaces. As I described in Section 6, we are now able to investigate ultrafast dynamics at liquid interfaces using TR-HD-VSFG spectroscopy, as we study ultrafast dynamics in solution using time-resolved IR/Raman spectroscopy. However, it is just the beginning. Thinking of the broad application of time-resolved spectroscopy to the ultrafast dynamics in the condensed phase, it is obvious that there remain a vast number of unexplored problems of interfacial dynamics that await investigation. The key interest in this research direction is elucidating the difference in the dynamics between liquid interfaces and solutions. In particular, elucidation of the reaction dynamics at the liquid interface is very important. Our UV-excited TR-HD-VSFG experiments have revealed that the photochemical reaction of phenol is accelerated by a factor of ∼10⁴ at the air/water interface,¹⁰² and it is considered that the difference in reaction dynamics between interface and solution is not specific to phenol, but is rather general. The success of tracking ultrafast reactions using UV-excited HD-VSFG spectroscopy can open a new wave in interfacial research, i.e. the study of chemical reactions and reaction dynamics at liquid interfaces. I note that time-resolved absorption spectroscopy has been most extensively utilized for the studies of ultrafast chemical reactions in solution. In this regard, time-resolved ESFG spectroscopy, in particular with heterodyne detection, will be a very powerful, general tool for studying chemical reactions proceeding at the interface, if it is realized.

Another promising direction is the extension of interface research from the gas/liquid interface to other interfaces, i.e. the solid/liquid, liquid/liquid, and solid/solid interfaces. Being different from the gas/liquid interface, these interfaces are sandwiched by 2 dense media, so they are called “buried interfaces.” Since it is very difficult to access buried interfaces and there are very few measurement methods for examining them, buried interfaces are left as unexplored research areas, and our understanding is very limited. Nevertheless, obtaining a molecular-level understanding of buried interfaces is highly desired because they are essential not only for fundamental science, but also in many practical and industrial applications such as batteries, devices, biocompatible polymers, etc. In particular, in situ measurements and analysis of buried interfaces are of critical importance for gaining a proper understanding of the molecular phenomena occurring there, enabling strategic designs and the realization of new functions and high efficiency at the interfaces, which is critical for the realization of a sustainable society. Second-order nonlinear spectroscopy, and, more generally, even-order nonlinear spectroscopy, is interface-selective and can access buried interfaces as long as one of the phases of the interface is transparent for the light used for the measurements, although actual measurements are technically difficult and elaborate, compared with the measurements of gas/liquid interfaces. We recently carried out a series of HD-VSFG studies on silica/water interfaces, which is one of the most fundamental solid/liquid interfaces.^114–116 At the silica/water interface, there is an equilibrium of SiOH ⇄ SiO^– + H⁺ and hence the silica surface is negatively charged, so the electric double layer (EDL) is formed at the interface. Taking full advantage of HD-VSFG spectroscopy, a clear molecular-level picture of the EDL has been obtained, and the pH-dependent change in the water structures in the contact Stern layer and diffused Gouy–Chapman layer has been clarified.¹¹⁶ Furthermore, we demonstrated that the in situ HD-VSFG measurement of electrode interfaces relevant to battery research is possible.^117,118 Although these are only a few examples of possible applications of HD-VSFG for buried interfaces, HD-VSFG has already shown its power in disentangling the complex spectra of buried interfaces to provide clear molecular-level pictures. For the study of buried interfaces, I consider that fourth-order Raman spectroscopy has the potential to become a versatile, powerful tool because it can be performed using only short pulses in the visible region.^{26,27,119,120} I also note that ultrafast dynamics at buried interfaces are completely unknown.

In the coming years, backed up by the continuous advances in technology, we are expected to examine weaker interfacial signals more often in attempting to obtain detailed information about interfaces, by leaving the strong OH stretch band. The interface selectively of second-order nonlinear spectroscopy, more generally, the interface selectivity of even-order nonlinear spectroscopy, is provided by the principle that the relevant nonlinear polarization is generated only in the region in which the inversion symmetry is broken, such as interfaces, under the dipole approximation. In other words, the interface selectivity is not guaranteed when the dipole approximation is not held. For VSFG spectroscopy, we reported several cases in which the dipole approximation is broken and the signal is generated not with the dipole mechanism, but with the quadrupole mechanism, for the molecule having the inversion symmetry (benzene)^121,122 and for weak vibrational bands (bend vibrations of water).^123,124 When we examine such interfaces, we are required to apply interface-selective nonlinear spectroscopy with sufficient care. Interestingly, however, these recent studies showed that the quadruple mechanism can also provide interface selectivity based on a principle that is different from the dipole mechanism: the large gradient of the electric field at the interface generates quadruple components localized at the interface region, which can give rise to the interface-selective nonlinear signal under certain conditions.¹²² This mechanism may open a new possibility for interface-selective nonlinear spectroscopy and allow us to examine the interface in which interfacial molecules are randomly oriented, which is beyond the scope of the application of interface-selective nonlinear spectroscopy so far.

Interface-selective nonlinear spectroscopy has been greatly developed in the past decades. Nevertheless, the elucidation of the properties, structure, and dynamics at a variety of interfaces at the molecular level has just begun. As far as there are important unsolved problems, it is worth tackling them with hope and dreams regardless of the difficulties. We will be able to obtain a variety of unprecedented new knowledge about fundamental molecular phenomena by continuing efforts to develop new approaches and apply them to untackled fundamental problems.

Acknowledgments

I heartily thank all of my coworkers who studied liquid interfaces with me at RIKEN. I can only mention a few names in this article, but other people also greatly contributed to the research. Without their hard work, we could not have realized the research I wrote in this article. In addition, I express my special thanks to Prof. Shoichi Yamaguchi at Saitama University for his careful reading of this manuscript and for helping me to recall what we actually thought and did at each point. I also thank 2 great theoreticians, Prof. Akihiro Morita at Tohoku University and Prof. Tatsuya Ishiyama at University of Toyama, for their long collaboration in the study of the liquid interface. Lastly, I acknowledge JSPS KAKENHI, which supported our interface studies. In particular, I appreciate the members of the KAKENHI projects, Grant-in-Aid for Scientific Research on Priority Area “Molecular Science for Supra Functional Systems (2007–2012),” and Grant-in-Aid for Scientific Research on Innovative Area “Soft Molecular Systems (2013–2017),” who greatly inspired me.

Funding

The studies described in this article were supported by JSPS KAKENHI Grants Numbers 19205005, 19056009, 22655009, 24245006, 25104005, 18H03905, 18H05265, and 23H00292.

References

Tahei Tahara

Tahei Tahara received his DSc degree from the University of Tokyo in 1989. He became a research associate of the University of Tokyo in 1989 and then moved to the newly founded Kanagawa Academy of Science and Technology as a research associate in 1990. In 1995, he joined the Institute for Molecular Science as an associate professor and started his own research group. He moved to RIKEN as Chief Scientist in 2001 and has been the director of the Molecular Spectroscopy Laboratory since then. His research interests are ultrafast spectroscopy, interface-selective nonlinear spectroscopy, and single-molecule spectroscopy of complex molecular systems.

Author notes