Article pubs.acs.org/JPCA
Unidirectional Vibrational Energy Flow in Nitrobenzene 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: Experiments were performed on nitrobenzene liquid at ambient temperature to probe vibrational energy flow from the nitro group to the phenyl group and vice versa. The IR pump, Raman probe method was used. Quantum chemical calculations were used to sort the normal modes of nitrobenzene into three categories: phenyl modes, nitro modes, and global modes. IR wavelengths in the 2500−3500 cm−1 range were found that best produced excitations initially localized on nitro or phenyl. Pulses at 2880 cm−1 excited a nitro stretch combination band. Pulses at 3080 cm−1 excited a phenyl C−H stretch plus some nitro stretch. With nitro excitation there was no detectable energy transfer to phenyl. With phenyl excitation there was no direct transfer to nitro, but there was some transfer to global modes such as phenyl-nitro stretching, so some of the vibrational amplitude on phenyl moved onto nitro. Thus energy transfer from nitro to phenyl was absent, but there was weak energy transfer from phenyl to nitro. The experimental methods described here can be used to study vibrational energy flow from one part of a molecule to another, which could assist in the design of molecules for molecular electronics and phononics. The vibrational isolation of the nitro group when attached to a phenyl moiety suggests that strongly nonthermal reaction pathways may play an important role in impact initiation of energetic materials having peripheral nitro groups. can be flash-heated by a laser pulse.5,6 Another way is to input energy into a particular location with submolecular resolution using scanning tunneling microscope tips and femtosecond IR pulses.7 The method used here involves using pump and probe laser pulses that create and probe vibrational energy transport using excitations that have local-mode character.8,9 This method relies on the well-known ability of IR-Raman spectroscopy to probe states other than the one being pumped. Until recently, only the IR-Raman method had this capability, but there have been recent advances in 2DIR methods that involve two different wavelength IR pulses10,11 or relaxationassisted 2DIR12−15 that also permit ultrafast studies of point-topoint vibrational energy transport. Some applications where vibrational energy flow and vibrational energy localization come to the forefront are molecular electronics, where molecules are used as currentcarrying wires,16−18 nanoscale heat transfer where it is hoped that useful molecular thermal diodes may be developed someday,19 and impact initiation of energetic materials, where vibrational energy concentration could create hot spots that grow explosively,20,21 causing widespread chemical reactivity that looks quite different from ordinary runaway thermal decomposition,22,23 and which occurs even with lower impact velocities that input relatively little net energy.24 Owing to the
1. INTRODUCTION In this study, we introduce a method based on ultrafast IRRaman spectroscopy1,2 to probe vibrational energy transfer from one part of a molecule to another. Using nitrobenzene as an example, we probed energy transfer from the nitro groups to the phenyl groups and vice versa. The molecules are in the liquid state at ambient temperature. IR-Raman is a type of three-dimensional spectroscopy,1 where the three dimensions are pump wavenumber, probe wavenumber and time. We used two kinds of 2D projections of the 3D IR-Raman surface for this study. With IR-Raman excitation spectroscopy, we tuned the IR pump pulse wavelengths at a fixed time delay while probing the resultant vibrational populations to find pulses that best generated vibrational excitations initially located on nitro or phenyl. Then fixing the pump pulses at either of these selected locations, we swept the time delay to probe energy flow between nitro and phenyl. The restricted energy flow between them, and the complete lack of energy flow in the direction of nitro to phenyl, in other words a net unidirectional vibrational energy flow, was surprising. The usual way to treat vibrational energy flow from an initial state prepared by an optical pulse (bright state) is to describe how that state time evolves into the continuum of background states at similar energies.3 Another approach is the nanoscale thermal conduction approach4 popular in materials science, which focuses on vibrational energy flow from one location to another. This approach needs a method to input energy into a particular part of a molecule and probe a different part. One way to do this is to fabricate a material where molecules are bound to a surface, especially a metal surface or electrode that © 2013 American Chemical Society
Special Issue: Prof. John C. Wright Festschrift Received: December 27, 2012 Revised: February 12, 2013 Published: February 25, 2013 6066
dx.doi.org/10.1021/jp3127863 | J. Phys. Chem. A 2013, 117, 6066−6072
The Journal of Physical Chemistry A
Article
ambient temperature capillary microjet that produces a stream 50 μm in diameter. At every delay setting, we computed the areas of the antiStokes Ias(t) and Stokes Is(t) transitions by fitting each transition to a Voigt line shape. The Voigt is not intended as a preconceived statement about the line shape, it is just used to get a good fit and a reliable metric of the peak area. Then the instantaneous change in occupation number for mode i, Δni(t), was computed using the relation,
significance or potential future significance of these applications, it is important to understand how molecular structure affects vibrational energy flow to different parts of a molecule, with the idea of eventually learning how to design molecules that channel vibrational energy in desired ways. One issue that arises with the Raman probing technique is selection bias. We do not probe every vibrational mode, only those with sizable Raman cross sections. For the nitro groups, we probe both stretching transitions that together constitute the vast majority of energy in the nitro groups. But for the phenyl group we selectively probe only some of the transitions, and for these transitions we can quantitatively determine the energy resident in each mode. The aggregate of this energy is called the “observed energy”, and we will view this observed energy flow as representative of the total energy flow. Our group has previously used an ultrafast calorimetry method to prove that the observed energy is truly representative of the total energy flow in the case of benzene.25 In this method, benzene was spiked with a molecular thermometer, CCl4,2,26−29 which monitored the total energy dissipation from IR-pumped benzene.25 By subtracting the observed energy from this total energy, we could determine the time dependence of the unobserved energy and except at the earliest times, its time dependence was the same as the observed energy. Thus the modes we do observe in the phenyl groups should provide a relatively unbiased measure of the total energy content of the phenyl groups.25 Because these are molecules in liquid, there will be intermolecular energy transport. This results in an indirect flow of energy from nitro to phenyl or vice versa, because the specific input vibrational energy will eventually be thermalized, meaning the vibrational energy will end up uniformly distributed throughout the liquid. For this reason the focus of this study will be the first few picoseconds, before much intermolecular energy transfer occurs. At these shorter delay times the dominant energy flow pathways are from the vibrationally hot part of the molecule into the rest of the molecule.25 The intermolecular effects would not be an important consideration in applications that involve singlemolecule electronic or phononic devices, but we must account for them in our liquid-state experiments.
Δni(t ) =
Ias(t ) ⎛ υL − υi ⎞ ⎜ ⎟ Is(t ) ⎝ υL + υi ⎠
3
(1)
where νL is the laser frequency and νi are the transition frequencies, and eq 1 is valid in the limit of small Δni(t). The excited-state energy density (J/cm3) of mode i can then be computed using the relation, ΔEi(t ) = NAhυiΔni(t )10−3m
(2)
where NA is Avogadro’s number and m is the molarity of nitrobenzene. The normal modes of nitrobenzene were calculated by MP2 perturbation theory with 6-31G as the basis using the Gaussian 09 computational package. These calculations were used to support literature assignments of the vibrational spectra.36
3. RESULTS Figure 1 shows the nitrobenzene Stokes Raman spectrum obtained with 0.7 ps laser pulses (25 cm−1 resolution) and the
2. EXPERIMENTAL SECTION The IR-Raman method, where IR pump pulses are used to excite molecular vibrations in a liquid, and 532 nm visible probe pulses are used to acquire anti-Stokes and Stokes Raman spectra over a −3800 to +3800 cm−1 range at various delay times, has been described previously.25,30−32 Since our most recent publication in this area,33 we have improved the temporal and spectral response of the system by improved pulse spectral masking and pulse compression. The pump and probe pulse durations were reduced to ∼0.7 ps and their spectral widths reduced to 25 cm−1. The apparatus response, measured using nonlinear light scattering (NLS) in water,34,35 has been decreased to 0.95 ps fwhm. An important new feature of the experiment is the capability for 2D IR-Raman excitation spectroscopy, where Raman spectra were obtained at fixed time delays while the IR pump pulse wavelengths were scanned, using a set of computer-controlled motors. Nitrobenzene (>99%, Acros Organics) was used without additional purification, after we verified that recrystallizing a small sample resulted in no improvement in Raman spectral backgrounds. This liquid was flowed through a recirculating
Figure 1. (a) Raman spectrum of nitrobenzene with the most intense transitions assigned. The spectrum was obtained using the picosecond laser with 25 cm−1 spectral resolution. (b) Infrared (IR) spectrum of nitrobenzene.
IR spectrum (4 cm−1 resolution) obtained with a commercial FTIR. The transition assignments are based on the literature,36 confirmed by our MP2 calculations described in the Supporting Information. Figure 2 shows mode assignments and the characters based on our MP2 calculations for vibrations visible in our IR-Raman experiment. The other invisible modes are included in the online Supporting Information where they are numbered corresponding to previous assignments. We have grouped the 36 normal modes into three categories, 25 phenyl modes, and 2 nitro modes, defined as modes where the atomic displacements were predominantly on phenyl or nitro groups, and 9 global modes, defined as those where atomic displacements were substantial on both moieties. In online versions of this paper 6067
dx.doi.org/10.1021/jp3127863 | J. Phys. Chem. A 2013, 117, 6066−6072
The Journal of Physical Chemistry A
Article
beneath this 2D spectrum, inasmuch as it represents the Raman cross-section variation with wavenumber. The signals at νL + νIR along the diagonal arise from anti-Stokes scattering from resonant nitrobenzene transitions at νIR directly pumped by IR pulses, but in addition there are contributions from coherent artifacts created by NLS that also appear at νL + νIR.37 These two contributions have different time dependences; the resonant signals decay with the excited-state lifetime T1 and the NLS signals decay with the apparatus temporal response. Figure 3 shows which pump wavenumbers produced the most phenyl ring excitation, as judged by the intensity of the νCC transition, and which IR pump wavenumbers produced the most nitro excitations, as judged by the intensities of νNO1 and νNO2. We could use the results in Figure 3 to find the pump wavenumbers that best selectively produced nitro or phenyl excitations. However, it is better to use occupation number changes rather than anti-Stokes intensities, so in Figure 4 we
Figure 2. Normal modes of nitrobenzene, visible in our IR-Raman experiment, computed using MP2 perturbation theory, and their division into three categories: phenyl modes, nitro modes, and global modes.
where color is available, we have color-coded these groups as black (phenyl), green (nitro), and red (global). One helpful fact to remember is that νCC (1590 cm−1) is a phenyl ring stretch and νNO1 (1512 cm−1) and νNO2 (1335 cm−1) are (respectively) antisymmetric and symmetric nitro stretches, as depicted in Figure 2. These assignments can be used for initial assessments of energy localization or energy flow between nitro and phenyl. Figure 3 is a 2D IR-Raman excitation spectrum, where the Raman probe pulses were fixed at 1 ps delay and the IR Figure 4. Instantaneous occupation number changes Δn(t) at 1 ps delay, of νCC (a phenyl vibration) and νNO1 and νNO2 (nitro vibrations), as the IR pump pulses were tuned. For reference, the IR absorption spectrum is superimposed. With 2880 cm−1 pumping, only nitro vibrations νNO1 and νNO2 were excited. With 3080 cm−1 pumping, phenyl vibrations were excited and also some νNO1.
have plotted Δn(t) for νCC, νNO1, and νNO2 as a function of IR pump wavenumber. On the basis of Figure 4, we see that 2880 cm−1 pulses create only νNO1 and νNO2 excitations with no νCC. This makes sense if we assign the weak IR transition at 2880 cm−1 as the νNO1 + νNO2 combination band. We will use this 2880 cm−1 IR wavenumber for nitro pumping experiments. There is no ideal IR wavenumber to produce only phenyl excitations. With 3080 cm−1, close to the phenyl νCH stretch absorption, we created a substantial amount of νCC with no νNO2, but unfortunately there was some νNO1. Thus the phenyl excitation experiments with 3080 cm−1 pumping will have an initial population of νNO1, and we will need to take that into account. Figure 5 shows time-dependent anti-Stokes spectra with nitro or phenyl pumping. Note the nonlinear time axis that has more spectra in the −3 to +10 ps range than in the 30−100 ps range. The nitro pumping results in Figure 5a show that the number of daughter excitations created by relaxation of the initial nitro excitation are very limited, and in fact almost all of the observed energy is seen only in the νNO1 or νNO2 transitions. These results make it clear that when vibrational energy is deposited on the nitro group, it stays on the nitro group and does not transfer to phenyl before it decays into the bath. This is not the case for phenyl pumping, where the daughter vibrations are numerous, and at least some of the daughter excitations were global excitations such as νCN1.
Figure 3. Results from 2D IR-Raman excitation measurements of nitrobenzene. The IR pump pulses were scanned from 2500 to 3500 cm−1 and the anti-Stokes Raman spectrum (minus a thermal background) was monitored at 1 ps delay. The anti-Stokes intensities are proportional to the instantaneous vibrational occupation number change Δn(t), except on the diagonal (shaded region) where there are artifacts from nonlinear light scattering. The Stokes Raman spectrum, the vibrational assignments, and the dotted drop lines are presented for reference. These results show which pump wavenumbers produced the most phenyl ring excitation, as judged by the intensity of the νCC transition, and which IR pump wavenumbers produced the most nitro excitations, as judged by the intensities of νNO1 and νNO2.
excitation pulses were tuned from 2500 to 3500 cm−1. The antiStokes spectral intensity changes (relative to the ambient temperature background) were plotted versus IR wavenumber. Recall that the anti-Stokes intensity changes are products of occupation number changes Δn(t) and Raman cross sections. The Stokes spectrum with assignments from Figure 1a is shown 6068
dx.doi.org/10.1021/jp3127863 | J. Phys. Chem. A 2013, 117, 6066−6072
The Journal of Physical Chemistry A
Article
state transition at 2880 cm−1 has a small Raman cross-section. Because we do not have enough spectroscopic resolution to detect the anharmonic shifts, the combination excitation νNO1 + νNO2 appears as simultaneous excitation of both fundamentals, νNO1 near 1512 cm−1 and νNO2 near 1335 cm−1.39,40 When we see both νNO1 and νNO2 signals, we cannot tell with certainty whether they represent two excitations on one molecule or one each on two molecules. In Figure 6b, both νNO1 and νNO2 signals have a component of instantaneous excitation because both have significant populations at t = 0. They both rise to the same occupation number, ∼3%. We believe the slightly different risetimes are due to noise from the measurement and data analysis, so we attribute these shorter-time signals to combination-band excitations created by IR pulses. After ∼1 ps, the νNO2 excitation level does not fall as fast as νNO1, even though at longer times
