Article pubs.acs.org/JPCB
Controlling Vibrational Energy Flow in Liquid Alkylbenzenes Brandt C. Pein, Yuxiao Sun, and Dana D. Dlott* School of Chemical Sciences, University of Illinois at UrbanaChampaign, Urbana, Illinois 61801, United States S Supporting Information *
ABSTRACT: Ultrafast infrared (IR) Raman spectroscopy was used to study vibrational energy in ϕ−S alkylbenzenes, where ϕ = C6H5 and substituents S were CH3− (toluene), (CH3)2CH− (isopropylbenzene, IPB), or (CH3)3C− (tbutylbenzene, TBB). Using methods described previously,1 the normal modes were classified as phenyl (ϕ), substituent (S), or global (G). IR pulses were tuned to find conditions that maximized the localization of initial CH-stretch excitations on ϕ or S. Anti-Stokes Raman spectroscopy measured transient energy content of Raman-active S, ϕ, and G modes, to determine the rates of phenyl to substituent (Φ → S) or substituent to phenyl (S → Φ) transfer during the first few picoseconds, when energy transfer was mainly intramolecular. Since phenyl CHstretches were 90−130 cm−1 uphill in energy from substituent CH-stretches, of interest were S → Φ processes where molecular structure and local couplings were more important than energy differences. The Φ → S process efficiencies were small and about equal with all three substituents. The S → Φ transfer efficiencies could be increased by increasing substituent size. This was opposite to what would be predicted on the basis of the larger density of states of larger substituents, and it provides a path toward controlling forward-to-backward vibrational energy transfer ratios. The S → Φ transfer efficiency is understood as resulting from an increase in the local anharmonic couplings. A heavier substituent, when vibrating, transfers energy more effectively to the phenyl group. granddaughters, etc. These multitier models14 provide frameworks to understand how molecular structures influence the energy flow within and between each other, but the influence is exerted by the vibrational density of states (DOS), since the multitude of individual anharmonic coupling coefficients are usually not considered explicitly. A classic example of this type focused on vibronic emission of jet-cooled n-alkylbenzenes. Smalley and co-workers15−17 demonstrated that increasing the alkyl chain length accelerated energy loss from the phenyl group. However, this type of study did not take into account details of molecular structure and how the intramolecular anharmonic couplings depended on local structure and interatomic distances. This is a problem because, for example, energy transfer between two nearby moieties seems more likely than that between two distant moieties, even when the DOS greatly favors the distant moiety. The proximity of terminal methyl groups in IPB and TBB with each other and the phenyl could affect phenyl relaxation. Additionally, Smalley’s experiments were measuring intramolecular vibrational relaxation (IVR) occurring in an excited electronic state. Comparing IVR in an excited electronic state may not be necessarily comparable to that in the ground state. In previous work,1 we developed an experimental protocol to study S → Φ and Φ → S transfer in molecular liquids, using nitrobenzene as a model system. Generally speaking, a
1. INTRODUCTION Vibrational energy flow plays a significant role in chemical reactivity and in nanoscale thermal transfer. For example, development of nanophononic technologies such as molecular thermal diodes and molecular wires2−4 would be greatly assisted by a detailed description of vibrational energy flow within the relevant molecular structures, and a theoretical framework for controlling vibrational energy by engineering the molecular structures. Toward these ends, in this study, we used the IR-Raman method to measure vibrational energy in a series of structurally relevant monosubstituted benzenes. These molecules, of the form ϕ−S, where ϕ is the phenyl group and S is the substituent, were studied in the liquid state to determine the rates of S → Φ and Φ → S transfer. The IRRaman method uses anti-Stokes Raman spectroscopy to probe time-dependent vibrational populations following the creation of an initial or parent vibrational excitation by frequencytunable ultrashort IR pulses. Monosubstituted benzenes are a particularly useful motif for these types of studies because the phenyl vibrational spectra and dynamics do not change much with different substituents. Vibrational energy flow in benzene itself has been studied in detail,5−9 and the perturbative effects of monosubstitution have been investigated using the series of liquids C6H5−X (X = F, Cl, Br, I).10 The earliest studies of vibrational energy in molecules relied on formulations using the golden rule,11 where the parent loses vibrational amplitude to states nearby in energy,12 called daughters. Subsequent advances have the daughters in turn losing energy to successive tiers of states13 containing © 2013 American Chemical Society
Received: July 2, 2013 Revised: August 14, 2013 Published: August 19, 2013 10898
dx.doi.org/10.1021/jp406528u | J. Phys. Chem. B 2013, 117, 10898−10904
The Journal of Physical Chemistry B
Article
was 99.5%, TBB 99%, and IPB 98%. The liquids were flowed through a recirculating ambient temperature capillary microjet 50 μm in diameter. The IR pump and visible probe pulses were about 0.7 ps in duration with 35−50 μJ (depending on the wavelength) and 50 μJ, respectively. The pump and probe pulses were focused to ∼60−70 and ∼60 μm diameter, respectively. Raman spectroscopy was performed using a spectrograph with a notch filter that simultaneously acquired Stokes and anti-Stokes spectra. Due to the spectral widths of the picosecond probe pulses, the Raman spectral resolution was 25 cm−1, which in many cases was greater than the natural linewidths. By combining the two types of Raman spectra, we could quantitatively determine the energy in each probed mode, as described previously.1 The IR spectra were obtained by FTIR, with 4 cm−1 resolution, which revealed the natural linewidths. The normal modes of each molecule were calculated using MP2 perturbation theory with the 6-31G basis set using the Gaussian 09 computational package. The resulting normal mode vectors and frequencies were used in conjunction with literature assignments21−24 of the vibrational spectra, to assign and classify each vibration as substituent, phenyl, or global.
complete description of this process would require a detailed knowledge of the potential energy surface of the molecular Hamiltonian and surrounding bath,18−20 and detailed measurements of the transient energy content in every normal mode. Lacking this, we developed a simpler method. Using quantum chemistry and literature vibrational assignments,21−24 we classified every mode as either phenyl (ϕ), substituent (S), or global (G), as explained in the Supporting Information. We then found an IR pulse wavelength in the CH-stretching region that maximized the localization of the initial vibrational excitation onto ϕ or S. With each of these IR pulses, we measured the energies in the phenyl, substituent, and global modes that had larger Raman cross sections and then computed the aggregate energy in each class of modes. We could then watch a phenyl excitation transfer energy to substituent modes and vice versa. Admittedly, this method suffers from the possibility of selection bias, since we observe only the modes with reasonably large Raman cross sections. However, in previous studies of benzene or d6-benzene, where 33 or 40% of the vibrations were directly observed, and where a simultaneous measurement of the total vibrational energy could be made using an external molecular thermometer, we showed that the energy content of the modes we did observe was entirely representative of the full energy content of the benzene molecules.5,7 We take this as compelling evidence that, in our studies of monosubstituted benzenes, the observed vibrations are representative of all vibrations. Because the molecules were in the liquid state, in addition to solvent-assisted IVR processes, there will also be vibrational energy transfer (VET) processes that result in energy flow from the parent or daughter vibrations into vibrations of adjacent molecules or into the bath. In order to focus on the effects of molecular structure and intramolecular couplings, and to better model molecular electronic devices where molecular components are largely isolated from their surroundings except at point contacts, here we will concentrate primarily on the first couple of picoseconds where the S → Φ and Φ → S vibrational energy transfer is primarily IVR.5,10,25 It is these shorter-time IVR processes that represent energy bursts flowing from one spatial location to another on the same molecule. As time progresses and successive tiers of the vibrational manifold become excited through VET, the liquid progresses toward thermalization. In the thermalization process, the initially localized vibrational energy will be spread out among all the liquid’s molecules and all the molecules’ atoms. In our nitrobenzene study, 1 we made a surprising observation. We found that phenyl-to-nitro Φ → S flow was inefficient but that nitro-to-phenyl S → Φ flow was entirely nonexistent. Thus, nitrobenzene acts as a vibrational energy diode, albeit not a very good or useful one. These results were clearly at odds by what would be predicted by simple DOS considerations, or by the overall downhill energetics of phenylto-nitro energy transfer. Instead, the observed behavior showed that substituted benzenes could demonstrate hierarchical mode−mode couplings that can effectively shut off S → Φ or hinder Φ → S energy flow. Thus, in the present study, we will pay particular attention to processes that deviate from what DOS considerations alone would predict.
3. RESULTS A. IR and Raman Spectra and Assignments. IR and Stokes-Raman spectra of the three alkylbenzene liquids are shown in Figure 1a−c. The transitions in the Raman spectra designated by assignments were those intense enough to be studied by transient anti-Stokes techniques. Smaller peaks, shoulders, and combination bands that were too weak to observe were not assigned. A full list of assignments and experimental and computed frequencies is given in the Supporting Information. Toluene has 39 modes of which we observed 11, IPB has 57 with 14 observed, and TBB has 66 with 15 observed. In toluene, ∼25% of the phenyl and ∼30% of the methyl and global modes were observed. With IPB, ∼30% of the phenyl, ∼30% of the propyl, and ∼20% of the global modes were observed. For TBB, ∼35% of the phenyl, ∼20% of the butyl, and ∼20% of the global were observed. On the basis of our previous study of benzene, we believe that each class of mode in each liquid was well represented. To clarify our labeling scheme, some mode descriptions are explained: νCH-phen (∼3050 cm−1) is the phenyl C−H stretch, νCC (∼1600 cm−1) is the phenyl C−C stretch with some ring C−H bend character, junction (jnct) modes (∼1200 and ∼700−800 cm−1) are global modes with displacements involving the substituent to phenyl C−C stretch, γCH (∼1150−1200 cm−1) are phenyl C−H bends, βCH (∼1000 cm−1) is ring breathing, and Φdef (∼650 cm−1) is a phenyl ring C−C−C deformation. The substituent C−H stretches (∼2950−3000 cm−1) and bends (∼1350−1450 cm−1) are labeled as νCH−X and γX where X = methyl for toluene, isopropyl for IPB, and t-butyl for TBB. There are C−C stretches within the substituents of IPB and TBB called (X)vCC (∼900 cm−1). Modes labeled tB-umb and me-umb are umbrella bending motions in TBB and toluene, respectively. Note that the me-umb mode is a CH-bending mode whereas the tB-umb mode is a C−(CH3) bending mode. These modes are analogous if we view the methyl groups in TBB as single unified atoms. Other notable modes include global wagging and rocking motions of the ring/substituent, and are labeled accordingly.
2. EXPERIMENTAL SECTION A. IR-Raman. The IR-Raman technique was described in detail previously.5,26−28 The liquids were supplied by Aldrich and used without additional purification. The purity of toluene 10899
dx.doi.org/10.1021/jp406528u | J. Phys. Chem. B 2013, 117, 10898−10904
The Journal of Physical Chemistry B
Article
Figure 1. Stokes-Raman and IR absorption spectra of (a) toluene, (b) isopropylbenzene (IPB), and (c) tertbutylbenzene (TBB). Raman spectra were acquired using picosecond pulses giving a resolution of 25 cm−1. IR spectra were acquired using FTIR with 4 cm−1 resolution.
Figure 2. Two-dimensional IR-Raman excitation spectra (top) and Stokes-Raman spectra (bottom) with the most intense transitions assigned, of (a) toluene, (b) isopropylbenzene (IPB), and (c) tertbutylbenzene (TBB). The dashed box indicates the tuning range of the IR excitation pulses that created parent excitations. The red dashed lines indicate which daughter excitations were populated to the greatest extent.
B. 2D Excitation Spectroscopy. Figure 2 shows 2D IRRaman excitation spectra for the three liquids. The waterfall plots are 2D slices of the full 3D spectra, and were obtained by frequency-scanning the IR pulses through the CH-stretch absorption regions (which were slightly different for each liquid) while monitoring the anti-Stokes spectra at a delay time of 1 ps. The excitation spectra were anti-Stokes difference spectra with the ambient temperature contributions subtracted off. Spectra at 1 ps are indicative of the excitations generated immediately at the end of each IR pulse. For reference, below each waterfall plot, we also show the annotated Stokes spectra. The parent excitations generated by IR pumping were observed along the diagonal. However, along the diagonal, there were also coherent artifacts whose amplitudes varied with IR wavenumber, that were created by nonlinear light scattering.28,29 The parent signals decayed with the excitedstate lifetimes T1, and the artifacts decayed with the apparatus temporal response (1 ps. The insets show the rise of daughter excitations during the first 10 ps. Stokes Raman spectra are displayed to help gauge relative Raman cross sections of each mode. Related transient IR-Raman spectra of IPB and TBB are shown in the Supporting Information.
little phenyl excitation, but phenyl pumping was not as selective, since phenyl pumping also generated substituent excitations. For IPB, phenyl pumping used 3060 cm−1 and substituent pumping used 2950 cm−1. As shown in Figure 3b, for IPB, both phenyl and substituent pumping were highly selective. For TBB, phenyl pumping used 3050 cm−1 and substituent pumping used 2960 cm−1. As shown in Figure 3c, phenyl pumping generated little substituent excitation and substituent pumping generated little phenyl excitation, but no matter where the IR pulses were tuned, a considerable amount of global excitations jnct2 and tB-umb were produced. Thus, in TBB, ϕ → S and S → ϕ energy transfer will always occur against a background of global excitations. As this discussion makes it clear, we could not achieve complete specificity of phenyl or substituent pumping with toluene or TBB, but for simplicity, in the next section, we will nevertheless refer to phenyl or substituent pumping, keeping in mind the specificity caveats discussed here. C. Anti-Stokes Transients. Figure 4 shows the results with toluene, of time-dependent IR-Raman experiments using either phenyl or the substituent (methyl) pumping pulses. The analogous results for IPB and TBB are shown in the Supporting Information. Note the nonlinear time axis on the right edge, which emphasizes the shorter time delays, the dashed lines indicating the pump pulse wavenumbers, and the insets focusing on daughter vibrations
