# How to determine conformation using IR

In my advisor's Ph.D. research, she used IR spectroscopy to (help) determine the conformation of a compound of interest (its structure is known.)

In organic chemistry, I learned about using IR as a way to identify functional groups (and in conjunction with HNMR, CNMR, & mass spec, I was able to determine the structure of a compound.)

How can one determine conformation using IR? What are the limitations of this? I feel like it could get messy really quickly.

• You maybe need to specify the type of compounds you're interested in. I've seen this kind of thing for large macromolecules, but not in small molecule elucidation – NotEvans. Jul 17 '17 at 16:31
• Assuming your advisor performed gas-phase IR spectroscopy, she could have determined the conformers based on predictions (e.g. DFT) for a known chemical structure. The intensities of the spectral lines are very sensitive to the dipole moments along the 3 principal axes of the molecule and they are typically different for different conformers and relatively easy to calculate. – Paul Jul 17 '17 at 20:23
• Another option is possible when she performed rotationally resolved spectroscopy (this is less likely). When you selectively replace each element in a molecule by a heavier isotope (e.g. replace H by D and $^{12}$C by $^{13}$C, the rotational constants change and by determining the constants from the spectrum you can work out the coordinates in the center-of-mass frame of the substituted atom (using the so-called Kraitchman Equations). – Paul Jul 17 '17 at 20:26
• I believe you're referring to this thesis. She uses vibrationally and rotationally resolved electronic spectroscopy. I don't know much about this so I'm not gonna answer, but the information is there if anyone wants to provide an answer. Feels kind of creepy I could find that using just your name and the question details. Haha :) – jheindel Jul 19 '17 at 4:08
• @jheindel I have to admit, that's pretty impressive. Blast my unusual name! – Melanie Shebel Jul 20 '17 at 2:43

This is perhaps my biased view as a theorist, but I've collaborated quite a bit with a condensed-phase 2D-IR group and a little with a UV resonance Raman group, so I have some understanding of the experimental perspective as well.

For anything outside of the simplest models, like a 2- or 3-state system you can scribble on an envelope, or the simplest molecules that would display configurational dynamics, like short peptides or single amino acids, experiment strongly benefits from computation. Have fun searching the literature for tables of IR peak frequencies, looking at the assigned normal modes, (which are sometimes compound motions), and then seeing if any of those match your hopefully very similar structure. If the molecular structures are known, then for simple cases a few of the conformers can be made by hand using your favorite computer program or a partially-automated conformer search, have their geometries optimized, and harmonic vibrational frequencies calculated to at least qualitative accuracy in an afternoon. Then it's a matter of clicking around in a visualization tool to see which interesting peaks correspond to which normal modes.

If populated conformers are harder to find, such as in systems that are larger with many rotatable bonds or have substantial librational motion, an exhaustive conformer search may be too expensive in terms of computer time, and running molecular dynamics trajectories with some sort of post-trajectory clustering analysis is an alternative, since IR and Raman spectra can be generated from trajectories via velocity, dipole, and polarizability autocorrelation functions.

## Example

Here is an example where a lithium cation is interacting with L-proline in two ways: 1. with the backbone-bound carboxylic acid, and 2. with the ring. Perhaps this is cheating, because it isn't a single molecule, it's two different coordination modes, but you would see something similar with an intramolecular conformational change or rearrangement such as keto-enol tautomerization.

Here are their respective vibrational spectra within the harmonic approximation.

Already without any assistance, we can see huge changes outside of the fingerprint region. This is similar to how hydrogen bonding in water changes the IR spectra when moving from molecules to the bulk. The three most notable changes going from backbone- to ring- bound are

1. a blueshift (increase in energy) of the carbonyl peak, moving it from 1760 to 1852 wavenumbers,

2. the intensity increase of the ring $\ce{C-H}$ stretches below 3000 wavenumbers, and

3. a few changes above 3500 wavenumbers.

1 is probably due to charge transfer into an antibonding orbital localized on the carbonyl group. However, it's also strongly coupled to the $\ce{-OH}$, so interactions there will have an indirect effect. I'm less sure about the $\ce{C-H}$ stretches, but they belong to the three hydrogens closest to the $\ce{Li+}$ (7, 11, and 14), and Mulliken charges don't support a CT idea, which isn't surprising and doesn't rule it out. For number 3, the $\ce{-OH}$ stretch moves from 3686 to 3439. This redshift is probably a result of the blueshift to the coupled mode; assuming that the total vibrational energy between conformers is similar ($E_{\text{ZPE}} = 90.100$ vs. $\pu{90.182 kcal/mol}$ here), vibrational energy tends to be redistributed evenly and in ways that make physical sense; see here for a discussion about how this simplified two-mode coupling works which might also be applicable here.

Frequencies are in wavenumbers (cm$^{-1}$) and intensities are in units of km/mol. Modes below 1200 wavenumbers are commented out.

$$\begin{array}{rrrr}% \hline% \text{backbone frequencies} & \text{backbone intensities} & \text{ring frequencies} & \text{ring intensities} \\ \hline% %% 30.91 & 2.37 & 62.75 & 4.92 \\ %% 88.36 & 6.77 & 91.13 & 2.54 \\ %% 168.98 & 24.56 & 146.67 & 15.42 \\ %% 200.22 & 35.17 & 239.15 & 53.54 \\ %% 301.33 & 5.59 & 299.49 & 10.23 \\ %% 323.29 & 67.62 & 399.53 & 47.70 \\ %% 393.10 & 18.41 & 436.61 & 77.33 \\ %% 423.78 & 49.21 & 454.79 & 10.29 \\ %% 485.16 & 69.90 & 517.44 & 4.46 \\ %% 534.48 & 56.37 & 541.24 & 57.13 \\ %% 565.71 & 45.28 & 582.68 & 14.18 \\ %% 625.82 & 27.87 & 622.71 & 9.03 \\ %% 663.98 & 12.16 & 703.37 & 26.78 \\ %% 754.48 & 22.99 & 758.60 & 12.72 \\ %% 802.57 & 16.27 & 808.16 & 50.74 \\ %% 842.01 & 5.53 & 823.92 & 48.36 \\ %% 858.14 & 14.85 & 863.84 & 7.69 \\ %% 875.87 & 27.62 & 875.21 & 13.13 \\ %% 897.13 & 13.53 & 912.31 & 2.55 \\ %% 919.58 & 0.12 & 928.34 & 12.11 \\ %% 966.10 & 12.84 & 952.79 & 7.90 \\ %% 1007.42 & 0.12 & 1017.17 & 0.53 \\ %% 1034.74 & 23.77 & 1056.15 & 1.11 \\ %% 1070.65 & 26.45 & 1116.38 & 40.56 \\ %% 1154.24 & 13.04 & 1119.11 & 9.28 \\ %% 1161.61 & 3.40 & 1148.10 & 7.38 \\ %% 1181.00 & 74.31 & 1173.28 & 7.56 \\ 1213.61 & 4.92 & 1203.94 & 6.81 \\ 1222.70 & 2.73 & 1228.83 & 4.07 \\ 1263.58 & 152.20 & 1261.21 & 39.44 \\ 1279.48 & 0.73 & 1272.66 & 12.99 \\ 1314.79 & 1.01 & 1321.29 & 57.22 \\ 1328.39 & 58.17 & 1342.94 & 314.05 \\ 1366.92 & 47.01 & 1365.77 & 74.02 \\ 1390.24 & 127.09 & 1402.33 & 19.65 \\ 1424.08 & 13.57 & 1522.07 & 16.68 \\ 1432.59 & 5.88 & 1561.70 & 0.40 \\ 1460.15 & 3.30 & 1572.71 & 24.88 \\ 1760.04 & 284.35 & 1851.86 & 270.80 \\ 2976.69 & 25.42 & 2607.42 & 324.10 \\ 2983.82 & 7.16 & 2806.11 & 147.60 \\ 2993.71 & 4.29 & 2900.28 & 54.24 \\ 3000.08 & 8.94 & 3024.05 & 5.13 \\ 3039.46 & 13.36 & 3027.04 & 2.76 \\ 3063.00 & 5.22 & 3061.92 & 2.04 \\ 3075.69 & 4.87 & 3074.89 & 3.80 \\ 3477.77 & 29.07 & 3438.98 & 235.23 \\ 3685.91 & 93.01 & 3527.08 & 36.70 \\% \hline% \end{array}%$$

All calculations were performed using B97-D/def2-SVP. See source for XYZ structures.