- 1 List of participants
- 2 Background
- 3 Some calculations on a real system: (dmpe)Cl3TiEt
- 4 Other C-H→M interactions
- 5 Possible questions for discussion
- 6 Relevant references
List of participants
Carlos Martín Fernández, KU Leuven.
Ferran Feixas, Universitat de Girona
Henry Rzepa, Imperial College London
The term agostic interaction was proposed by Brookhart and Green  to "refer specifically to situations in which a hydrogen atom is covalently bonded simultaneously to both a carbon atom and to a transition metal atom." The word agostic is derived from the Greek word αγοστος which occurs in Homer and translates as to "clasp" or "hold to oneself".
The agostic interaction can be seen as an intermediate between a compound with a metal-alkyl bond and an alkylidene hydride, as shown in Fig. 1. This is one of the main reasons behind the fact that this kind of interaction appears numerous times in reaction mechanisms involving C-H bond activation.
Some characteristics of the agostic interactions are:
- Low 1H-NMR field shifts.
- C-H bond lengthening.
- Reduced νC-H stretching frequencies.
- Short M-H distance.
Usually, agostic interactions are classified as α, β, γ... depending on the length of the chain of C atoms connecting the metal and the interacting CH group, as shown in Fig. 2.
Although there is a large amount of effort in trying to understand this kind of interaction/bond better, it is very challenging to do so. It is so weak that depending on the method you might or might not get a good geometry or wavefunction that will have the proper characteristics of an agostic interaction. And there has been some debate also on how to characterize different but similar CH→M interactions, such as agostic, anagostic, hydrogen bonds... Doing a careful review of the literature on this topic is, then, a challenge on its own.
Some calculations on a real system: (dmpe)Cl3TiEt
Carlos (talk): I have run some calculations on model complex (dmpe)Cl3TiEt (Fig. 3), which is a classic β-agostic complex. The structure was optimized at M06/def2TZVPP level of theory, and all calculations have been run with the same geometry/functional/basis set unless noted otherwise.
- Geometries: From the structure it is clear that the agostic interaction causes a number of distortions. Firstly, the Ti-C-C angle is only 84.6°, which is very distorted from an expected angle of ∼109° of a tetrahedral C atom. Also the C-Hagost length is 1.123 Å, as compared to the 1.089 Å of the C-Hnon-agost. Finally, the distance between the Ti atom and the Hagost is 2.083 Å.
- Frequencies: Another charateristic of the agostic interaction is that the C-Hagost stretching mode will be red shifted. And we can clearly see from our calculations that νC-Hagost is shifted to ∼2755 cm-1, while the other C-Hnon-agost show a symmetric and antisymmetric mode at 3072 and 3140 cm-1, respectively.
Agostic C-H stretching mode
- Electron density shifts: I have also calculated the intramolecular electron density shift as proposed by Sánchez-Sanz and co-workers (Fig. 4), and we can see the results in Fig. 5. It is clear from this figure that there is an increase in electronic density between the metal atom and the Hagost one. Also, we can see that the dmpe ligand does not seem to play an important role for the agostic interaction. Using this approach, we can also estimate the interaction energy, and we obtain an Eint ≈ 23 kJ/mol. This partition scheme is interesting because it can be used to compute interaction energies from the isodesmic reaction. One of the issues related to this partition, though, is that it is not feasible for α-agostic interactions.
- QTAIM: Usually a bond (even a weak one, such as a hydrogen bond) is associated with the existence of a bond path, and we can explain its nature by looking at certain properties at the bond critical point (BCP). However, the agostic interaction is tricky to study with this technique. The absence of a BCP in some agostic interactions has been explained before (look at the review by Scherer and McGrady, particularly Fig. 10) by the merging of the RCP and BCP, and they can be borderline systems of a bond catastrophe. In fact, if we compare the M06/def2TZVPP and B3LYP/def2TZVPP//M06/def2TZVPP we can see that in the M06 case we obtain no BCP, but we obtain it using B3LYP functional (Fig. 6). In any case, we can see a clear difference in densities with the C-H bonds for both cases. Using the M06 density, for C-Hagost we obtain ρBCP = 0.254 au, while for C-Hnon-agost we obtain ρBCP = 0.279 au. With the B3LYP density we also obtain a BCP between Ti and H, which presents ρBCP = 0.035 au. If we look at the electron density contours (Fig. 7), we can see that the overall description of the density is quite similar in both cases, with the exception of the appearance or not of a BCP. Are we again in a situation where Bond Critical Points are not really "critical" for agostic bonding?
- NCI: If we look at the NCI calculation, we can see clearly (Fig. 8) that the agostic interaction appears as a clear attractive interaction between the metal and H. However, it is also close to a repulsive area probably from the C atom. Note that this repulsive area only appears on the side of the alkyl chain, but if one rotates the molecule and looks at it from the metal "point of view", it looks attractive. Is this an artifact of the representation method? Also, it is noticeable (from the color scheme) that the agostic interaction looks as strong (blue) as the P-Ti bonds... All the connections with the Ti atom have been removed for the sake of clarity
- --Henry (talk) 08:48, 12 August 2017 (CEST)The interactive NCI depiction looks great! I notice that it loads a cube density file and then produces an NCI iso-surface. The cube file is 4.3 Mbyte in size (1.6 Mbyte if gzipped, which I think JSmol supports?). One can select just the isosurface and express that as a .jvxl file, which I find is just 20 Kbyte in size. For those whose bandwidth is perhaps a bit slower, this latter file would load far faster than the cube file. And one can then use much higher resolution cube files to generate the .jvxl. Just a suggestion!
- --CarlosMF (talk) 19:09, 16 August 2017 (CEST) I tried to do it using an .jvxl file, but I could not manage and that is why I used the .cube... If someone manages to work it out with the .jvxl (s)he is most welcome to edit the page!
Fig. 8 3D plot of the NCI calculation
- NBO: I have run some NBO calculations, in which the agostic interaction can be seen by looking at different parameters. Firstly, by looking at the occupation numbers of the NBOs it is clear that the σC-Hagost has a smaller occupation than the σC-Hnon-agost. And if we look at the perturbative treatment of the orbital interactions (that can account for processes like charge transfer), we can see the main transfers from the BD(C-Hagost) orbital are to the BD*(Ti-P) or LP*(Ti) orbitals.
- CAUTION! By running calculations with different basis sets and different versions of the NBO program, one can get a similar result in terms of the different populations of the σC-H orbitals, but the E(2) analysis can be very different, particularly in determining the nature of the acceptor orbital. See table below
|Comparison of NBO calculations with M06 functional||NBOs|
|NBO version||Basis set||Donor (occupation)||Acceptor (occupation)||Agostic E(2) in kcal/mol||BD (C-H) → LP* (Ti)|
|NBO 3.1||6-31G||BD C-H (1.88301)||LP* Ti (0.18384)||40.55|
|6-31+G(d,p)||BD C-H (1.83360)||LP* Ti (0.21532)||57.14|
|6-311G||BD C-H (1.90861)||BD* Ti-P (0.12239)||14.20|
|6-311+G(d,p)||BD C-H (1.82713)||LP* Ti (0.22099)||64.25|
|def2TZVPP||BD C-H (1.82269)||LP* Ti (0.21635)||59.34||BD (C-H) → BD* (Ti-P)|
|NBO 6.0||6-31G||BD C-H (1.92007)||BD* Ti-P (0.11394)||12.26|
|6-31+G(d,p)||BD C-H (1.87547)||BD* Ti-P (0.13104)||25.32|
|6-311G||BD C-H (1.90962)||BD* Ti-P (0.12639)||14.64|
|6-311+G(d,p)||BD C-H (1.87877)||BD* Ti-P (0.13149)||24.71|
|def2TZVPP||BD C-H (1.90922)||BD* Ti-P (0.12868)||15.05|
- ELF: In Fig. 10 we can see the representation of the ELF calculation. There is a clear difference between the Hagost and the Hnon-agost. Actually, it has been reported  that the ELF signature of agostic hydrogens is: 1. Metal QTAIM basin contribution to V(H,..)) population which ranges between 0.02 to 0.3 e. 2. The covariance matrix element of the V(H,..) and C(M) population < -0.02.
- I would be happy if someone could run an ELF calculation on this complex, as it might help in giving a more complete picture of the interaction.
- --Henry (talk) 09:36, 10 August 2017 (CEST) Re: ELF, do you have the .wfn file available for analysis?
- --CarlosMF (talk) 14:34, 10 August 2017 (CEST) ELF calculations are being run by prof. B. Silvi, so hopefully he will post something about the results he obtains shortly.
- --Henry (talk) 08:33, 11 August 2017 (CEST)The ELF does indeed show an interesting difference. Has the .wfn file been uploaded to the Wiki perchance? I ask because it would be good to have the volume and integrations of the basins as a quantitive measure of the difference.
- --CarlosMF (talk) 14:13, 11 August 2017 (CEST) The wfn file at the M06/def2TZVPP level has been uploaded to http://dipc.ehu.es/bondslam/images/6/66/Ti_agost_wavefun.wfn in case anyone wants to run calculations on it
- --CarlosMF (talk) 14:57, 11 August 2017 (CEST) The output from the ELF is in http://dipc.ehu.es/bondslam/images/3/36/TiEtCl3dmpe_popfile.pop Since I do not know much about the interpretation of the ELF results, I very much appreciate all possible comments.
- --Henry (talk) 07:21, 12 August 2017 (CEST)The ELF analysis of the wavefunction follows that of the NBO partition. Hagostic has a basin integration of 1.95e and a volume of 50.0 (the smaller volume is very apparent from the figure above). The non-agostic hydrogens have integrations of 2.00 and volumes of 76 - 77.5.
- --Henry (talk) 07:21, 12 August 2017 (CEST)Bernard (Silvi): a question for you. The .wfn file provided for us was calculated using the Def2-TZVPP basis, which includes f-functions. I used TopMod09 for the analysis, which thinks this wavefunction has a pseudopotential. Might it be that the f-functions are confusing it? The results above used a WFN file recalculated using the 6-311G(d,p) basis instead, which does not use a pseudopotential for the Ti. The population file provided by Bernard (I think using a more recent version of Topmod?) returns values of 1.95 and 2.05 for the populations and 48/64-67 for the volumes, showing the results are mildly basis-set dependent.
Other C-H→M interactions
Over the last years, some differences have been noted for certain C-H→M contacts, prompting new (improved?) definitions of the agostic interaction nad/or the other interactions. For instance, in their 2007 review, Brookhart, Green and Parkin make a difference between "agostic" and "anagostic" interactions:
|3c2e interaction||largely electrostatic interaction|
|d(M-H) ≈ 1.8 - 2.3 Å||d(M-H) ≈ 2.3 - 2.9 Å|
|M-C-H ≈ 90 - 140°||M-C-H ≈ 110 - 170°|
|δH upfield of uncoordinated CH||δH downfield of uncoordinated CH|
But in a rather recent paper it has been argued that there are two kinds of anagostic interactions (H-anagostic and C-anagostic), and also it is argued that the traditional agostic (what they call η2(C,H) form) can be "regarded as a balanced symmetrical hybrid of the basic H- and C-anagostic forms".
I have calculated the NMR shieldings for the complex, using the same method as you do, but adding CDCl3 as solvent to make it more realistic. The chemical shift of the agostic proton is ~4.73 ppm, whilst the other two protons attached to this carbon come at 1.15 and 1.47 ppm. Thus the agostic proton is downfield with respect to the non-agostic protons. According to the classification above (which I think is too simplistic), this is an anagostic interaction?
Possible questions for discussion
- Agostic "bond" vs. agostic "interaction"
- Should only CH→M interactions be considered agostic? Or also for other cases with different σ bonds (BH, CC, SiH...)?
- Should (and how) agostic interactions be taken into account for "electron counting" in organometallic complexes?
- How to quantify the strength of an agostic interaction?
- Brookhart, M.; Green, M. L. H. Carbon-hydrogen-transition metal bonds. J. Organomet. Chem., 1983, 250, 395-408. DOI:10.1016/0022-328X(83)85065-7
- Clot E.; Eisenstein O. (2004) Agostic Interactions from a Computational Perspective: One Name, Many Interpretations. In: Principles and Applications of Density Functional Theory in Inorganic Chemistry II. Structure and Bonding, vol 113. Springer, Berlin, Heidelberg. DOI:10.1007/b97940
- Lein, M. Characterization of agostic interactions in theory and computation. Coord. Chem. Rev., 2009, 253, 625-634. DOI:10.1016/j.ccr.2008.07.007
- Brookhart, M.; Green, M. L. H.; Parkin, G. Agostic interactions in transition metal compounds. PNAS, 2007, 104 (17), 6908-6914. DOI:10.1073/pnas.0610747104
- Scherer, W.; McGrady, G. S. Agostic interactions in d0 Metal Alkyl Complexes. Angew. Chem. Int. Ed., 2004, 43, 1782-1806. DOI:10.1002/anie.200200548
- Sánchez-Sanz, G.; Trujillo, C.; Alkorta, I.; Elguero, J. Electron density shift description of non-bonding intramolecular interactions. Comp. and Theor. Chem., 2012, 991 (1), 124-133. DOI:10.1016/j.comptc.2012.04.007
- Lane, J. R.; Contreras-García, J.; Piquemal, J-P.; Miller, B. J.; Kjaergaard, H. G. Are Bond Critical Points Really Critical for Hydrogen Bonding?. J. Chem. Theory Comput., 2013, 9 (8), 3263–3266.DOI:10.1021/ct400420r
- Lepetit, C.; Poater, J.; Esmail Alikhani, M.; Silvi, B.; Canac, Y.; Contreras-García, J.; Solà, M.; Chauvin, R. The Missing Entry in the Agostic-Anagostic Series: Rh(I)-η1-C Interactions in P(CH)P Pincer Complexes. Inorg. Chem., 2015, 54(6), 2960-2969. DOI:10.1021/acs.inorgchem.5b00069
- E-L. Zins, B. Silvi, M.E. Alikhani, Phys. Chem. Chem. Phys, 2015, 17, 9258-9281, DOI:10.1039/c4cp05728g
- For wavefunction and ELF analysis, see DOI:10.14469/hpc/2889
- H. S. Rzepa, 2017, DOI:cbp3
- H. S. Rzepa, 2017, DOI:cbqz
- H. S. Rzepa, DOI:cbr6
- H. S. Rzepa, DOI:cbr5