Chemistry 324 - Spring 2008

NMR - General

As we discussed in class, always use the very best data tables you can find to predict chemical shifts and coupling constant, but remember that spectral data + assignments for a structurally analogous compound are even better than data tables. Some places you can look for spectral data on specific compounds:

• The same paper. Chemists tend to make a series of analogous molecules, replacing a methyl with a phenyl here, or a proton with a methoxy group there. If one of these molecules possesses an easily interpreted spectrum, you might be able to apply this same interpretation to your molecule’s spectra.
• Books. The Reed library contains several compilations of NMR (1H and 13C) and IR spectra (and more). Two favorites of mine are the various Aldrich Libraries (also available online) and “Tables of Spectra Data for Structure Determination of Organic Compounds,” by Pretsch et al. (QC462.85 .T313 1983).
• Online Tables. The Sigma-Aldrich site lets you look at high quality spectra for many of the compounds they sell. You get the spectra, but no interpretation. Another excellent site that actually provides assignments is maintained by Prof. Hans Reich, U. Wisconsin-Madison. I have listed links to four of his data tables, but his site contains much, much more:
  • ¹H chemical shifts
  • HH coupling constants
  • ¹³C chemical shifts
  • CH coupling constants

I have also added some other useful links for organic chemists to the site’s side bar. Check them out before the Qual.

Comments on specific data for compound 3

¹H NMR

Aromatics. It is hard, but not impossible, to predict the chemical shifts of the aromatic protons in compound 3 using data tables. However, you probably noticed that the aromatic protons produce two patterns of 2H and 3H; the first pattern must be either ortho or meta H, but which? Although the Supporting Information does not contain the spectrum of 3, it does contain spectra for structurally related compounds. These hard-to-read spectra indicate that the 2H pattern is something like a doublet, i.e., it should be assigned to the ortho H, so you might reason that this holds true for 3 as well.

Funky J values. Don’t be overwhelmed by minor discrepancies in coupling constants. Here are two that you noticed:

• The -CH₂-CH= protons show vicinal couplings of 7.4, 7.2, and 6.9 Hz. Shouldn’t these all be the same?
• Apparently one of the =CH₂ protons shows a geminal coupling of 1.3 Hz (H trans to -CH=), but its geminal neighbor (cis to -CH=) does not. Shouldn’t the same coupling appear in both patterns?

Many factors can affect the appearance of a coupling pattern (see below). If you prefer to skip my list, just remember this: coupling constants are experimental measurements. All experimental measurements are subject to random uncertainties, operator error/bias, instrumental error/bias, and strange coincidences.

Now for my list of “maybes”:

• Uncertainty in the data. The NMR spectrum is digitized, not continuous, i.e., it is a collection of signal intensities measured at discrete frequencies (data points). If the points of measurement are 0.2 Hz apart (a typical value), then the location of peaks and the spacings between peaks will be uncertain by at least 0.2 Hz (peak positions listed in a “peak pick” give the locations of these data points; they do not indicate the actual uncertainty in peak position).
• Operator bias. The operator determines the digitization of the spectrum, i.e., she sets the number of data points and the width of the spectral window. If the operator sets up an unusually broad window, or a relatively small number of data points (goes faster, takes up less space on disk), then the points of measurement will be separated by more than 0.2 Hz and peak position will be even more uncertain. This will also lower spectral resolution and peaks that are close together may appear as a single peak.
• Spin relaxation processes guarantee that signals appear as “broad” peaks, not lines. Relaxation rates vary from one proton to the next, so peak widths vary across a spectrum. In addition, operators typically introduce more broadening when they process raw NMR data in order to improve signal-to-noise. These factors combine so that signals that are close together (typically within 0.5-1 Hz) may appear as a single peak.
• Accidents happen. The -CH= proton appears to be coupled to the two CH₂ protons with coupling constants of 7.4 and 6.9 Hz (these are the constants that appear in the CH₂ signals). However, these values are simultaneously uncertain (see above) and similar. Consequently, the -CH= may appear as a triplet instead of a doublet of doublets with an “average” coupling constant of 7.2 Hz.

¹³C NMR

Alkene, aromatic, and carbonyl carbon nuclei are all strongly deshielded, but the degree of deshielding is very sensitive to molecular structure. Looking at Hans Reich’s data tables, I see the following:

• CO2H at 178 ppm in HOC(Me2)CO2H
• CH=CH2 at 138 and 116 ppm, respectively, in CH2=CHCH2CMe3

Did you notice that the compounds I cited contain branching patterns and/or functional groups that are similar to compound 3?

Compound 3 produces three signals, 179.0, 140.2, and 120.4 ppm, that seem to match up with these observations. This leaves the signals at 131.7, 128.4, 128.2, and 125.5 to be assigned to the aromatic carbons. If I could see the intensities of these signals I might be able to differentiate the ortho/meta signals (these would be doubly tall) from the para/ipso signals (these would be short). Also, I might assign the shortest peak to the ipso carbon because carbons without any attached protons tend to produce weak signals. But I don’t have intensity data, and the chemical shifts are very close together, so I will just have to let this part rest without making any definite assignments.

High-resolution MS

This is a technique for determining a molecule’s molecular formula. It relies on the fact that molecules that appear to generate identical molecular ions at low resolution, e.g. H2C=CH2, CO, and N2 all produce molecular ions at m/e 28, can be distinguished under high resolution. However, at high resolution, you could probably distinguish the molecular ions of these molecules because:

• H2C=CH2 m/e 28.0313
• CO m/e 27.9949
• N2 m/e 28.0062

High-resolution MS data is easily obtained on molecules of modest size. Also, m/e are easily predicted using online calculators (like this one; scroll to bottom of page to find it).

“Old school” chemists don’t like to rely solely on HRMS because the ability to characterize one ion doesn’t say anything about the purity of a sample.

Elemental analysis

Sometimes called combustion analysis or CH analysis, this technique gives you the mass percentage of different elements in a substance. The data are always reported as percentages calculated for a given formula followed by percentages observed. For compound 3, we are told: “Anal. Calcd for C11H12O3: C, 68.74; H, 6.29. Found: C, 68.48; H, 6.21.“

It is a trivial matter to check the Calcd quantities. For example, compound 3’s %C should be (11 C * atomic mass C)/(molecular weight C11H12O3) = (11*12.011)/192.215 = 0.6874 or 68.74%. %H should be (3 H * atomic mass H)/molecular weight C11H12O3) = (3 * 1.008)/192.215 = 0.0629 or 6.29%.

Something that is far more difficult to generate (and usually not all that useful) is a list of all possible substances that might give elemental analyses consistent with the observed analysis. For example, a dimer (or trimer, or …) of compound 3 must give exactly the same analysis as compound 3. On the other hand, the methyl ester of compound 3 (C12H14O3) should give C, 69.88; H, 6.84; these values are clearly inconsistent with the experimentally observed values of C 68.48; H, 6.21.