The Optics Laboratory
Group of Hans Hallen and Russell Philbrick, Physics Department, North Carolina State University
Our resonance Raman studies  involve true resonance, as the incident wavelength is tuned directly over the absorption line. The resonance arises as an increase in the intensity of the Raman signal near an absorption. The classic explanation for this effect is that the photon spends more time at the atom near an absorption, when the index of refraction is large so the propagation rate of light slow, so it has more time to cause a vibration to happen. As in any Raman process, if a photon initiates a vibration, it looses energy equal to that input to the vibration, and this amount is quantized by quantum mechanics, so discrete peaks at particular energy loss (dependent upon the material and its bonds) appear. The various vibration energies thus measured are often used as a fingerprint to identify the material and its state. We find that this picture can't be correct, as particular choice of an absorption feature can force the interaction time to be extremely short (which we know because it doesn't have time to share energy with its neighbors during the process), and the resonance factor is still extremely large. This is best understood with an example, and we will give several after a short methods summary.
There are three reasons why we use resonance Raman in the deep UV excitation range. The first is that many more molecules have absorptions in the UV than elsewhere The second is to reduce background, and is driven by the small signal strength of Raman signals. It is very unlikely that a photon from the incoming light will set off a vibration and loose energy, perhaps one in a million. The third is that the molecular nature of the process means a dipole-like interaction of the incoming and outgoing fields, which makes the nonresonant Raman signal proportional to the frequency of the excitation light to the fourth power. The shorter wavelengths of the UV have higher frequencies, so higher signal. Other competing factors, such as scattered incident light (which can work its way through the spectrometer and become background) can be reduced by proper sample positioning and a good spectrometer, but fluorescence form the sample is harder to eliminate. Due to the increased absorption in the UV, more fluorescence is expected there, and is typically why Raman is done at the longer wavelengths, particularly for complex molecules such as biomolecules. Figure 1 shows why this is not important for deep enough UV excitation. In particular, the fluorescent emission, which normally appears at a lower energy that the excitation by a roughly fixed amount, gets 'hung up' at ~280 nm, so ceases to be a background as the Raman lines (must) continue to track at fixed energies below the excitation. This 'sticking' has to do with electron de-excitation dynamics. Electrons very quickly thermalize to the bottom of a band via non-optical energy loss, then take longer to emit light and return to the band they belong in. There are enough overlapping bands and band minima to provide a range of emission energies for lower energy excitation, but once the electrons are energized to a high enough level, then there are not as many possible band minima, and they loose energy. The fact that this appears to happen at about 4.4 eV (280 nm) is empirical, and is expected to vary by 10's of nm with different materials.
Figure 1. Raman spectra at different excitation show how the Raman (at fixed energy separation from the excitation), 'moves away' from the fluorescence background at lower energies. A few major Raman lines are indicated for each excitation .
Benzene is a simple molecule, six carbons in a ring with a hydrogen attached to each. Among its absorption lines is one near 260 nm that is phonon-allowed (symmetry forbidden). We measure resonance Raman across several of the absorption plus phonon creation peaks shown in Fig. 2, the largest of which features are separated by the energy of the 990 cm-1 phonon: the ring stretch mode that is strongly coupled to this 'ring back bond breaking' absorption.
Figure 2. The top frame shows three absorption spectra, two for liquid benzene, as used in these experiments, and one high resolution absorption spectra of vapor phase (isolated molecule) benzene. The lower frame zooms in and uses purple lines to identify the wavelengths used for Fig. 3 , .
Figure 3. The Raman spectra at the indicated excitation wavelengths change dramatically in strength over this 2 nm excitation wavelength range. This is much too narrow for a solid or liquid state process.
Figure 4. Raman peak strength at two vibration lines is compared to the isolated molecule (vapor) absorption spectra to show the correlation.
Figure 5. Top or Left: Profile of the Raman peak strength integrated over several vibration lines shows dramatic variation with some very narrow and other broader (and weaker) features. Bottom or Right: Raman peak strength from above is compared to the isolated molecule (vapor) absorption spectra to show the correlation. The largest Raman gain obvious at the left is cut off so that other features are visible. Now the Raman measurement has higher resolution .
Toluene is similar to benzene, it is obtained by replacing one of the hydrogens in benzene with a methyl group. Thus, is still has the aromatic ring and the equivalent phonon-allowed absorption near 260 nm excitation. To check the generality of this phenomenon, we measure resonance Raman across several the absorption peaks shown at the top of Fig. 6. The effect is seen, lower frame of Fig. 6, and (this used the wider bandwidth laser) the range of wavelengths shown is less than the corresponding benzene figure. Again, the spectra are weak at the excitations at the ends of the range used, and peak where the isolated molecule absorption does. Clearly the same effect is occurring. Several times we have seen the behavior of the blue curve at 266.64 nm -- when the energy of excitation is slightly above the peak absorption, many additional Raman like lines fill the spectral region. We do not have an explanation at this time.
Figure 6. The top frame shows the vapor phase (isolated molecule) absorption spectra, with vertical lines indicating the excitation energies used in the lower part. The lower frame contains Raman spectra at those excitation wavelengths .
The resonance Raman of ice shows many standard features, which makes it a great comparison sample to benzene and toluene. The Raman spectra over a broad excitation range are shown in Fig. 7. The spectra are offset for clarity. Qualitatively, the 3 bumped Raman spectra varies little in shape, but there are higher and lower Raman gains in this region. To better study these, we integrate the spectra, correct for the inherent frequency to the fourth power mentioned above, and plot the result as a function of excitation wavelength in Fig. 8. This shows a generally increasing signal level with some features on top of it. The general increase is a typical pre-resonant Raman effect, over a broad wavelength range, following a Lorentzian line shape, as is shown fit in the figure. In this case, it is towards an actual symmetry allowed absorption at ~150 nm. The smaller features on top of it are emphasized by subtracting the pre-resonance, and are shown in Fig. 9. These peaks correspond to ice absorptions that are 'final state effects.' In particular, when ice is driven to the higher state with the absorption, it has a larger dipole moment per molecule. This adds energy to the system unless the neighbors can re-arrange and use their intrinsic dipole moments to reduce the impact of the dipole moment. The net lowering of the energy of the excited state means that a lower energy (longer wavelength) photon can excite the molecule. These absorptions, at longer wavelengths than the ~150 nm absorption, are what drive the observed band-like resonance highlighted in Fig. 9. Note that these absorptions rely on neighbor interactions, so provide a counterpoint to the above observations on symmetry forbidden absorptions. The resonant gain profiles are much narrower than the large pre-resonant slope, suggesting a higher level of coherence around these interactions.
Figure 7. Raman spectra of ice at several wavelengths in the UV and visible, offset for clarity .
Figure 8. The integral under the Raman signal in Fig. 7, corrected for the intrinsic frequency to the fourth power factor, as a function of the excitation wavelength reveals two contributions to the resonance gain.
Figure 9. After subtraction of the pre-resonance of the ~150 nm absorption, a band-like resonance associated with final state effect absorptions is visible.
At resonance, we are actually looking at the energies of excited state vibrations, not those of the ground state as in off-resonance Raman. These differ because the bonding is different -- the excited state may have a broken bond. The result of this is that some vibrations get stiffer, some less stiff, and some are unchanged , . Also, the vibrations most impacted by the bond that has changed will be likely impacted also strongly in energy, and often are the most highly enhanced. In fact, while multiplets and combinations are very rare in nonresonant Raman, they are very common in the modes strongly enhanced by the resonance. Figure 10 shows the energy change in several benzene bonds as resonance is approach. The figure illustrates the shift from the ground state to the excited state vibration manifold. The ν2 vibration has only a small energy shift and a different functional form with energy separation than the ν10 or the very large ν9 related vibration modes.
Figure 10. Vibration energies from the peak position of the Raman spectra are tracked for several modes and combination mades as resonance is approached.
The benzene absorption shown in Fig. 2 has more narrow peaks than near 260 nm. There is a family of features replicated by the ring stretch mode of benzene of ~990 wavenumbers. Essentially, for each replicate, an additional ring stretch phonon is created along with those discussed above. With Fig. 11 we verify that the same effects occur at these energies as well. The spectra show that although the coupling to the modes changes the relative strengths of the lines, indicating a difference in the coupling between the vibration modes and the excitation cased by the particular absorption.
Figure 11. Raman spectra near other, related absorption features show the narrow, strong resonance and changes in resonant coupling.
The resonant gain per molecule can be higher than these data suggest due to absorption changes with wavelength, i.e., the number of molecules observed decreases as the absorption increases. For narrow spectral regions such as we have been showing, the relevant liquid absorption does not change significantly, so the effect is minimal. The absorption can be different for the incoming excitations and exciting Raman shifted photons, due to the energy change. Figure 12 shows the geometry. The signal decreases exponentially along the path. Note that resonance Raman is always associated with absorption, since the excitation has to be near an absorption for the resonance condition. At resonance in a bulk sample, the absorption reduces the signal while the resonance increases it; the net gain in real signal is minimal , . Where resonance really helps is in trace analysis in a non-absorbing medium, or for nanoscale Raman .
Figure 12. A schematic drawing of the input and exit light paths illustrates the distances along which exponential decay of the light occurs.
More info is in the papers.
NC State University | Physics | Optics Home
Copyright ©2000-2017, Hallen Laboratory, NCSU, Raleigh, NC. www.physics.ncsu.edu/optics
Comments or questions? Hans_Hallen@ncsu.edu