[spirochete]

An ester has a carbonyl stretch in the IR of about 1740 cm^-1. For a ketone, it is about 1715. The explanation for this perplexes me when I also consider other experimental data from different kinds of spectroscopy. The standard explanation given here is inductive electron withdrawal being more important than resonance.

But in carbon NMR, the carbonyl carbon of an ester shows up at a much lower ppm than the carbonyl carbon of an ester. And in hydrogen NMR, alpha hydrogens of an ester show up at slightly lower ppm than alpha hydrogens of a ketone.

Also, during chemical reactions, a ketone carbonyl tends to be a stronger electrophile than an ester carbonyl. This part seems easier to explain, because the energy of the LUMO is clearly a major issue here.

I understand that IR vs NMR is an apples to oranges comparison, because IR involves vibrational frequency and NMR involves the difference between how different nuclei experience a magnetic field. But I still don't understand what factors are causing it to work out like this.

[wildfyr]

@Corribus, this one is right up your alley

[Babcock_Hall]

This is not entirely within my comfort zone, but I am happy to kick around ideas.  @OP, What would you predict about the C=O bond lengths in ketones versus esters and why?

[Corribus]

@spirochete

I think comparisons between FTIR and NMR are dangerous. They measure very different things.

Inductive effect explanation for FTiR makes sense: If you have an -OR group pulling electron density away from the carbonyl carbon, it will be more positively charged, which will make it bond more tightly to the carbonyl oxygen. Tighter bond = shorter bond length = higher vibrational frequency. Also note that esters don't really have stable resonance structures, so these can't significantly contribute to the electronic structure around the carbonyl. Compare for example carbonyl stretching frequency of undissociated and non-hydrogen-bonded carboxylic acid (kind of like an ester), around 1760 cm-1 to that of carboxylate ion, which has strong resonance contributions - asymmetric stretch at around 1650-1550 cm-1. In the latter case, resonance structures spread out the bond order, resulting in a longer bond (pair of bonds, actually) with lower vibrational frequency.

[Babcock_Hall]

The comparison between the stretching frequencies of the C=O group in esters versus amides is also suggestive of the importance of resonance delocalization in some instances.

[Meter]

@spirochete

I think comparisons between FTIR and NMR are dangerous. They measure very different things.

Inductive effect explanation for FTiR makes sense: If you have an -OR group pulling electron density away from the carbonyl carbon, it will be more positively charged, which will make it bond more tightly to the carbonyl oxygen. Tighter bond = shorter bond length = higher vibrational frequency. Also note that esters don't really have stable resonance structures, so these can't significantly contribute to the electronic structure around the carbonyl. Compare for example carbonyl stretching frequency of undissociated and non-hydrogen-bonded carboxylic acid (kind of like an ester), around 1760 cm-1 to that of carboxylate ion, which has strong resonance contributions - asymmetric stretch at around 1650-1550 cm-1. In the latter case, resonance structures spread out the bond order, resulting in a longer bond (pair of bonds, actually) with lower vibrational frequency.

Hi, I hope it is okay I ask here.

I noticed that certain IR spectra of conjugated carboxylic acids have slightly different -OH stretches than in non-conjugated versions (like benzoic acid vs. acetic acid, for example). Is this a pattern that repeats?

[Corribus]

It seems reasonable to predict that any carbonyl that is included in an extended conjugation will have lowered bond order, longer bond length, and vibrational transitions shifted to lower frequencies - all things equal of course. Mining FTIR data is challenging because FTIR spectra are very sensitive to to the method used, the matrix, etc. - and this is especially so for carboxylic acids because they dimerize when they are "pure" and dissociate in moisture. To really get an idea of how molecular structure influences the carbonyl resonance, you have to have spectral data acquired in dilute nonpolar solvent (or "solvent", such as a solid or oil mull). But I gave it a try: here are some comparisons that might be interesting to you. FYI, the carbonyl resonance of acetic acid at about 1715 cm-1 in dilute nonpolar media is usually taken as a good reference point.

Here are carbonyl stretching frequencies for molecules in which the carbonyl is conjugated to an aryl group - you can see they are all shifted to lower frequences relative to acetic acid benchmark. References are in [].

benzoic acid (carbon tet) 1695 cm-1 [NIST webbook]
para-toluic acid (carbon tet) 1691 cm-1 [NIST webbook]
ortho-toluic acid (solid kbr) 1678 cm-1  [https://doi.org/10.33805/2639-6734.106]
meta-toluic acid (solid kbr) 1689 cm-1  [forgot to write this down, but it was by the same authors]
naphthoic acid (split mull) 1679 cm-1 [NIST webbook]
methyl thiophene carboxylic acid (solid mull) 1669 cm-1 [NIST]

By contrast, just about every aliphatic carboxylic acid that I could find has a carbonyl stretch at around 1705-1720 cm-1. Of notable comparison to those listed above:

benzene acetic acid (solid mull) has a carbonyl stretch at 1712 cm-1 [NIST webbook]. I.e., introducing a single CH2 group between the benzene ring and the carboxylic acid disrupts the benzene-carboxylic acid conjugation and shifts the stretching frequency to higher frequencies due to loss of resonance effects.

The saturated and unsaturated C18 acids are revealing:

Dilute oleic acid (9-octenoic acid, 18:1), linoleic (18:2 cis-9,12), linolenic (18-3 cis-9,12,15), arachidonic ( 20:4 5,8,11,14) acids in CS2 all have almost identical carbonyl stretching frequencies at 1708 cm-1. Likewise, the fully saturated stearic acid (octanoic acid) also has a stretching frequency at  1708 cm-1. By contrast, 2-octenoic acid (both cis and trans) has a carbonyl stretching frequency at 1695 cm-1. I.e., regardless of how many double bonds the acid has, or where it is located, the C=0 stretching frequency is about the same, unless the double bond is immediately adjacent to the carbonyl, in which cases resonance contributions come into play. [Ref:Sinclair et al, JACS, 1952, 2579]

As a final example, consider these (all from NIST webbook):

4-chloro benzoic acid (oil mull) ~1680-1685 cm-1
3-chloro benzoic acid (m o mull) ~1695-1700 cm-1
2-chloro benzoic acid (m o mull) ~1685-1690 cm-1
(I had to do the chloro-substituted ones by eye, hence the uncertainty)

4-fluoro benzoic acid (solid mull) 1680 cm-1
3-fluoro benzoic acid (solid mull) 1691 cm-1

These are all molecules in which halogen-substituted aryl groups are conjugated to the carbonyl. In all cases resonance effects shift the carbonyl to lower frequencies compared to aliphatic carboxylic acid, as expected, but the amount of shift depends quite a bit on where the halogen is situated on the ring. I.e., the 3-substituted positions are shifted a little to higher frequencies compared to 2- and 4-substituted positions. To me, this suggests that inductive effects through the aryl group due to electron withdrawing substitutents can also influence the carbonyl vibrational frequency, and the amount of induction depends on the electronic structure of the benzene ring. It can't be a coincidence that if you draw the resonance structures out for benzoic acid, the partial-positive charge at the carbonyl carbon will be partially distributed on the 2- and 4- (and 6-) positions of the aryl ring - and halogen substitution at these positions gives the strongest down-frequency resonance shift of the carbonyl vibration. Conversely, substitution at the ortho position, which cannot be the recipient of distributed positive charge via resonance to the carbonyl carbon, results in a carbonyl stretch that is shifted slightly to higher frequencies compared to the benchmark benzoic acid. I.e., halogen substitution at the 2- and 4- positions significantly stabilizes resonance structures, leading to lower frequency shift of the carbonyl, whereas substitution at the 3-position destabilizes resonance structures, having the opposite effect. Just spit balling there but it's an interesting observation.

EDIT:
I found more reliable C=0 stretching frequencies for the chlorobenzoic acid (Ref: Lee et al. Spect. Chim. Acta A, 1996, 52(2) 173-184):

In Kbr pellet:
p-chlorobenzene, 1684 cm-1
o-chlorobenzene, 1696 cm-1
m-chlorobenzene, 1692 cm-1

[wildfyr]

You did a tremendous amount of work for a forum post Corribus!

FTIR really is very tricky to find reliable trends in, much more so than NMR.

[Babcock_Hall]

https://pubs.rsc.org/En/content/articlepdf/1947/tf/tf9474300158
http://dx.doi.org/10.3998/ark.5550190.0012.506

The first link is an old paper that has a few C=O bond distances and force constants.  The second link has a table with force constants and bond lengths for C=O and C=S bonds.

[spirochete]

Some good replies here. Unfortunately I posted this question at a time when I am very busy and can't reply to everyone.

I would assume IR is the best experimental data relating to C=O bond length*. Therefore, an ester should have a shorter C=O bond than a ketone. Without considering this experimental data, I might've wrongly assumed resonance would make the C=O bond length longer in an ester. Ultimately I would need to research the bond lengths. Also, researching accurate partial charges on the carbonyl carbon for various compounds would be interesting. I don't have access to the paper Babcock posted.

Long story short, I think I just had a tendency to over-emphasize the importance of resonance with oxygen (and sometimes nitrogen) lone pairs in neutral molecules. Especially with oxygen. Clearly with nitrogen lone pairs the electron donation is important, although I have read that the resonance argument even has some flaws with amides, although this is sort of cutting edge computational stuff that I don't understand. I also read somewhere reputable that the nitrogen in aniline is somewhat pyramidal, which is more easily determined experimentally.

When drawing the minor resonance contributors of an ester, it is interesting to think what the point is. There are some correlations, but they only work out in some cases.

*Edit: X-ray crystallography would be the best for bond lengths, assuming the solvent doesn't change it much, which I don't think it would.