I have solved a crystal structure of an enatiopure product. This product had 80 ee% based on HPLC equipped with a chiral column. I would like to make sure that the absolute structure was correct because the standard deviation of the Flack parameter was greater than 0.04, which I believe it needs to be below the range to be conclusive about the absolute structure. I may be wrong, but if a crystal contains an enantiopure product, the standard deviation can be tolerated up to 0.1 instead. Could the latter apply to my result to confirm the correctness of the absolute structure? Please refer to the information below:
Xray source: MoKa
Formula: C28 H21 N1 S1
R1/wR2 = 3.74/9.20 %
Completeness: 99.6 %
Flack x: 0.05(7)
Hooft y: 0.034(11)
Validation on Absolute Structure by Flack Parameter

 Trained Monkey
 Posts: 6
 Joined: 13 Aug 2019, 04:36

 Guru
 Posts: 33
 Joined: 05 Oct 2010, 06:09
Re: Validation on Absolute Structure by Flack Parameter
Two thoughts:
1. You do not have an enantiopure compound. An ee of 80% leaves a significant fraction of the other enantiomer. You will have to test multiple crystals to make sure they all represent the same absolute structure. There may even be crystals of the racemate in your batch, with another space group!
2. I personally prefer to look at the likelihood ratios given by PLATON rather than at “y” and its standard uncertainty. Check the linearity of the probability plots too. It is slightly worrying that y deviates three sigma from zero.
1. You do not have an enantiopure compound. An ee of 80% leaves a significant fraction of the other enantiomer. You will have to test multiple crystals to make sure they all represent the same absolute structure. There may even be crystals of the racemate in your batch, with another space group!
2. I personally prefer to look at the likelihood ratios given by PLATON rather than at “y” and its standard uncertainty. Check the linearity of the probability plots too. It is slightly worrying that y deviates three sigma from zero.

 Rotating Anode With Optics
 Posts: 269
 Joined: 26 Jan 2012, 16:04
Re: Validation on Absolute Structure by Flack Parameter
Another thought: With sulfur as heaviest element and MoKalpha radiation you are at the very limit of being able to determine an absolute structure. I have had much worse e.s.d.'s from datasets like that even from compounds that where synthesised starting from natural products, i.e. >99% enantiopure.

 Sealed Tube
 Posts: 28
 Joined: 26 Nov 2014, 13:05
Re: Validation on Absolute Structure by Flack Parameter
Given the Friedif value of 83 for C28H21NS for Mo radiation, it should be possible to obtain a sufficiently low esd on the Flack/Parsons/Hooft. See doi:10.1107/S0108767307002802 on the Friedif thing.
However, I agree with rwwh and Helge: 80 % ee is far off what I would call 'enantiopure'.
Surprisingly, the esd of Flack x (I guess it's Parsons z of SHELXL as it usually yields more significant esds than the postrefinement x') is lower than that of Hooft y. What's the coverage of Bijvoet pairs? Use the 'ByvoetPair' function of PLATON.
For the likelihood ratios: I don't think that they're useful here as there is no qualified preassumption to be made.
1) If you are sure that there can only be exactly one enantiomer present (it must be either one or the other enantiomer, but there can't be both present!), you can take the P2(true) value.
2) If it is possible that it is a 1:1 twin (perfect inversion twin between R and S), look at the P3(true)/P3(false) values. P3(ractwin) tells you the probability that you're dealing with a perfect inversion twin.
x', y and z don't rely on assumptions like those: they take the two cases above PLUS the possibility that you're dealing with any nonperfect twinning by inversion, e.g. 80 % / 20 %, into account. So, use those in this case!
As you already know, for z(u) (or x(u), x'(u), y(u), ...), you will require u < 0.1 for enantiopuresufficient inversiondistinguishing power and u < 0.04 for strong inversiondistinguishing power. The first one is adequate if you're sure that there can only be one enantiomer present (you can't in this case!), the second one is required when there are no preassumptions to be made. Then, you'll require z (or x, x', y, ...) < 3u. Else, you might want to refine as a racemic twin (TWIN and BASF z as a starting point; BASF should be greater than three times its esd in order to be of significance). See doi:10.1107/S0021889800007184 on the criteria for value/esd.
However, I agree with rwwh and Helge: 80 % ee is far off what I would call 'enantiopure'.
Surprisingly, the esd of Flack x (I guess it's Parsons z of SHELXL as it usually yields more significant esds than the postrefinement x') is lower than that of Hooft y. What's the coverage of Bijvoet pairs? Use the 'ByvoetPair' function of PLATON.
For the likelihood ratios: I don't think that they're useful here as there is no qualified preassumption to be made.
1) If you are sure that there can only be exactly one enantiomer present (it must be either one or the other enantiomer, but there can't be both present!), you can take the P2(true) value.
2) If it is possible that it is a 1:1 twin (perfect inversion twin between R and S), look at the P3(true)/P3(false) values. P3(ractwin) tells you the probability that you're dealing with a perfect inversion twin.
x', y and z don't rely on assumptions like those: they take the two cases above PLUS the possibility that you're dealing with any nonperfect twinning by inversion, e.g. 80 % / 20 %, into account. So, use those in this case!
As you already know, for z(u) (or x(u), x'(u), y(u), ...), you will require u < 0.1 for enantiopuresufficient inversiondistinguishing power and u < 0.04 for strong inversiondistinguishing power. The first one is adequate if you're sure that there can only be one enantiomer present (you can't in this case!), the second one is required when there are no preassumptions to be made. Then, you'll require z (or x, x', y, ...) < 3u. Else, you might want to refine as a racemic twin (TWIN and BASF z as a starting point; BASF should be greater than three times its esd in order to be of significance). See doi:10.1107/S0021889800007184 on the criteria for value/esd.