Skip to content

Explicit bonds between metal and hydrogen atoms must be kept #168

Closed

Description

@gblanke02

The current InChI releases including the development release (status version 1.07.4) delete any explicit terminal hydrogen atoms bound to a metal atom and keep the information in the /h layer.

Example: hydrido(dimethyl)iron: InChI=1/C2H7Fe/c1-3-2/h3H,1-2H3 with atom 1 and 2 as C, and atom 3 as Fe

To identify the stereochemistry of molecular inorganic compounds, any H atom directly bound to a metal atom must be kept, has to undergo the canonicalization process and must be listed in the connectivity string
Expected
InChI=1/C2H7Fe/c1-4(2)3/h1-2H3
With atom 1 and 2 as C, atom 3 as H and atom 4 as Fe. Note that the /h layer is adapted accordingly.

hydrido(dimethyl)iron
  ACCLDraw11182517382D

  4  3  0  0  0  0  0  0  0  0999 V2000
    6.0938   -6.6250    0.0000 Fe  0  0  0  0  0  0  0  0  0  0  0  0
    7.1166   -6.0344    0.0000 H   0  0  0  0  0  0  0  0  0  0  0  0
    5.0709   -6.0344    0.0000 C   0  0  0  0  0  0  0  0  0  0  0  0
    6.0938   -7.8061    0.0000 C   0  0  0  0  0  0  0  0  0  0  0  0
  1  2  1  0  0  0  0
  1  3  1  0  0  0  0
  1  4  1  0  0  0  0
M  END
iron_trihydride
  ACCLDraw11182517552D

  4  3  0  0  0  0  0  0  0  0999 V2000
    6.0938   -7.8061    0.0000 H   0  0  0  0  0  0  0  0  0  0  0  0
    5.0709   -6.0344    0.0000 H   0  0  0  0  0  0  0  0  0  0  0  0
    7.1166   -6.0344    0.0000 H   0  0  0  0  0  0  0  0  0  0  0  0
    6.0938   -6.6250    0.0000 Fe  0  0  0  0  0  0  0  0  0  0  0  0
  4  1  1  0  0  0  0
  4  2  1  0  0  0  0
  4  3  1  0  0  0  0
M  END

Note: In case of metal hydrides like iron trihydride FeH3, the current implementation is sufficient as the connectivity string is not created: InChI=1/FeH3/h1H3.

Activity

  1. gblanke02 commented on Nov 21, 2025

    @gblanke02
    ContributorAuthor

    Explicit bonds between one H atom and two metal atoms are already kept. B-H-B systems are handled as well. In both cases, the H atom gets an canonical atomnumber, is listed in the connectivity string and does not occur in the /h layer.

  2. schatzsc commented on Nov 27, 2025

    @schatzsc

    @gblanke02

    To identify the stereochemistry of molecular inorganic compounds, any H atom directly bound to a metal atom must be kept, has to undergo the canonicalization process and must be listed in the connectivity string Expected InChI=1/C2H6ClFe/c1-4(2)3/h1-2H3 With atom 1 and 2 as C, atom 3 as H and atom 4 as Fe. Note that the /h layer is adapted accordingly.

    The sum formula of your expected InChI example is not correct - with two methyl groups and a metal-coordinated hydrido ligand, it should be:

    InChI=1/C2H7Fe/c1-4(2)3/h1-2H3

    2x CH3 + 1x FeH = 6+1 = 7 H

    And where is the Cl in your example coming from???

  3. schatzsc commented on Nov 27, 2025

    @schatzsc

    @gblanke02

    Explicit bonds between one H atom and two metal atoms are already kept.

    It actually does not matter if the two "other" atoms are metal or not - any bridging H atoms (those with coordination number/"actual valence" larger than one) of the type X(µ-H)X' must be kept, regardless which elements X and X' are.

  4. gblanke02 commented on Nov 27, 2025

    @gblanke02
    ContributorAuthor
  5. gblanke02 commented on Nov 27, 2025

    @gblanke02
    ContributorAuthor
  6. gblanke02 commented on Jan 8, 2026

    @gblanke02
    ContributorAuthor

    Any bridging H atoms (those with coordination number/"actual valence" larger than one) of the type X(µ-H)X' are kept already, but the atom numbers are the highest in the list, i.e. the H atom numbers of the type "X(µ-H)X" are counted from the highest atom number of the non H-atoms and are not listed in the /h layer.
    To keep the approach equivalent for explicit H atoms bound to (one) metal atom they have to get the highest atom number positions. For our example above that means tah the explcit H atom gets the highest atom number:
    InChI=1/C2H7Fe/c1-3(2)4/h1-2H3

  7. gblanke02 commented on Jan 8, 2026

    @gblanke02
    ContributorAuthor

    The following 2 structures represent a Th complex where the two Th atoms are bridged by H atoms twice while each Th atom gets an additional H atom and 2 cyclopentadeiney rings.
    The original example has been found in Reaxys whie teh second one is drawn by using the haptic bonds of Biovia/Draw

    The 3rd example represents a Ru complex that has got H2 bound described by a haptic bond. The resulting INChI is calculated as : InChI=1B/H12O5Ru/c1-6(2,3,4,5)7-8-6/h1-5H2 with atom numbers 7 and 8 as explcit H atoms.

    inchi3 org - Biovia - haptic adapted.mol.txt
    inchi3 org.mol.txt
    Ru(OH2)5H2 - Hydrogen complex.mol.txt

    The following exmple contans metal-hydrogen atom bonds but handles all H atoms in the hydrogen layer.
    H2PtNH2.mol.txt

  8. schatzsc commented on Jan 8, 2026

    @schatzsc

    @gblanke02

    Is that the intended behaviour to keep the terminal hydrides in the /h layer?!?

    For H2PtNH2.mol.txt the InChI is

    InChI=1B/H5NPt/c1-2/h2H2,1H3

    I have no problem with that - but is there a consistent treatment, like all bridging H in the /c layer and all terminal H in the /h layer, regardless if they were explicitly defined in the molfile or implicitly calculated?

  9. schatzsc commented on Jan 8, 2026

    @schatzsc

    @gblanke02

    I also tried the ruthenium complex [Ru(H2)(H2O)5] which has the expected 12H count - but I'm a bit confused about the structure display in the lower right. Is it intentional that there are three white "spheres" shown - with the central one presumably the "* pseudo atom"?!? If that is indeed the case, it should be displayed differently from the hydrogen atoms.

    Also, something does not seem to be right with the /c layer:

    InChI=1B/H12O5Ru/c1-6(2,3,4,5)7-8-6/h1-5H2

    It appears to me when you select the "canonical index" that the node number #6 is incorrectly assigned to the "* pseudo atom" and not one of the oxygen atoms, among which one remains un-labelled.

    The canonical labels are: O1, O2, O3, O4, Ru5, H7, H8

    Label #6 is on the "third white sphere" and one O has no canonical label at all - looks like an indexing problem

    Image
  10. gblanke02 commented on Jan 9, 2026

    @gblanke02
    ContributorAuthor

    @gblanke02

    Is that the intended behaviour to keep the terminal hydrides in the /h layer?!?

    For H2PtNH2.mol.txt the InChI is

    InChI=1B/H5NPt/c1-2/h2H2,1H3

    I have no problem with that - but is there a consistent treatment, like all bridging H in the /c layer and all terminal H in the /h layer, regardless if they were explicitly defined in the molfile or implicitly calculated?

    @schatzsc
    The explicit H atoms directly bound to a metal atoms must be represented in the connectivity layer because we need the atom numbers for the representation of the stereochemistry. For the numbering we should follow the H atom number ssystem of the type "X(µ-H)X".

  11. gblanke02 commented on Jan 9, 2026

    @gblanke02
    ContributorAuthor

    @gblanke02

    I also tried the ruthenium complex [Ru(H2)(H2O)5] which has the expected 12H count - but I'm a bit confused about the structure display in the lower right. Is it intentional that there are three white "spheres" shown - with the central one presumably the "* pseudo atom"?!? If that is indeed the case, it should be displayed differently from the hydrogen atoms.

    Also, something does not seem to be right with the /c layer:

    InChI=1B/H12O5Ru/c1-6(2,3,4,5)7-8-6/h1-5H2

    It appears to me when you select the "canonical index" that the node number #6 is incorrectly assigned to the "* pseudo atom" and not one of the oxygen atoms, among which one remains un-labelled.

    The canonical labels are: O1, O2, O3, O4, Ru5, H7, H8

    Label #6 is on the "third white sphere" and one O has no canonical label at all - looks like an indexing problem
    Image

    @schatzsc
    Ferrocene shows a similar issue in the structure display tool. The display in the structure display tools is based on information of the auxinfo. The auxinfo keeps the original * atoms that are taken out during the determination of the InChI in case of haptic bonds. The auxinfo is not adapted accordingly (it has to represent the original input structure). What you see in the "discrepancy" Between the original structure and the related InChI. I have created a seaparte issue for that.

  12. schatzsc commented on Jan 9, 2026

    @schatzsc

    @gblanke02

    The explicit H atoms directly bound to a metal atoms must be represented in the connectivity layer because we need the atom numbers for the representation of the stereochemistry. For the numbering we should follow the H atom number ssystem of the type "X(µ-H)X".

    Ok, so in other words the following InChI is not the proper required format:

    InChI=1B/H5NPt/c1-2/h2H2,1H3

    Instead, it should be something like

    InChI=1B/H5NPt/c1-2(3,4)/h1H3

    But will H order higher or lower than N and Pt?!?

    That's also something I still find confusing - canonical numbering in the InChI is by alphabetic order of the element systems, not the atomic number?!?

  13. gblanke02 commented on Jan 20, 2026

    @gblanke02
    ContributorAuthor

    The InChI uses the HIll formula (https://en.wikipedia.org/wiki/Chemical_formula#Hill_system).

    At the molecular inorganics working group meeting of 19-Jan-2026 we decided to handle the explicit H atoms directly bound to a metal atom in the connectivity layer and not in the hydrogen layer “/h”. Because InChI does not provide any rule for implicit H atoms at metal atoms the H atoms must be drawn explicitly. In consistency with any H atoms in bridge positions like in the "X(µ-H)X" type mentioned in the comments above the explicit H atoms directly bound to metal atoms are counted on from the highest non-Hydrogen atom number within the InChI and represented in the /c and not in the /h layer.

    That means for the NH5PT example we get
    InChI=1B/H5NPt/c1-2(3,4)/h1H3 with 3 and 4 as the 2 explicit H atoms directly linked with the PT atom.

  14. self-assigned this
    on Jan 27, 2026
  15. nnuk commented on Jan 27, 2026

    @nnuk
    Collaborator

    According to the rules described above, with this commit, the issue has been resolved and can be tested on the InChI Web Demo.

  16. added a commit that references this issue on Oct 16, 2026
Sign up for free to join this conversation on GitHub. Already have an account? Sign in to comment

Metadata

Metadata

Assignees

Labels

molecular inorganicsto be discussedThis issue needs further discussions to come up to a decision or solution

Type

No type

Projects

No projects

    Milestone

    No milestone

    Relationships

    None yet

    Development

    No branches or pull requests

    Issue actions