Generalize template for orthonormalize

[EricaCMitchell]

Description

Currently, all orthonormalize_* functions require the input vectors and overlap to have the same type. In the case of complex types, the input vectors would be complex but the overlap would be real.

Details

• Change template to have different types for madness::Function and madness::Tensor

[robertjharrison]

Why would the overlap be real? Hermitian maybe.

…

On Tue, Oct 22, 2024 at 2:32 PM Erica Mitchell ***@***.***> wrote: Description Currently, all orthonormalize_* functions require the input vectors and overlap to have the same type. In the case of complex types, the input vectors would be complex but the overlap would be real. Details - Change template to have different types for madness::Function and madness::Tensor ------------------------------ You can view, comment on, or merge this pull request online at: #555 Commit Summary - 8e4d98a <8e4d98a> Allow different type for ovlp File Changes (1 file <https://github.com/m-a-d-n-e-s-s/madness/pull/555/files>) - *M* src/madness/mra/vmra.h <https://github.com/m-a-d-n-e-s-s/madness/pull/555/files#diff-97f081c69164f087e3d9c9aa4aaa78ea6ec3b41575fb5668580b76260bd95fae> (48) Patch Links: - https://github.com/m-a-d-n-e-s-s/madness/pull/555.patch - https://github.com/m-a-d-n-e-s-s/madness/pull/555.diff — Reply to this email directly, view it on GitHub <#555>, or unsubscribe <https://github.com/notifications/unsubscribe-auth/ABZSAPLXHP3V2AWWQONN3QLZ42K3VAVCNFSM6AAAAABQND4CAKVHI2DSMVQWIX3LMV43ASLTON2WKOZSGYYDMMJYGQ2TSNY> . You are receiving this because you are subscribed to this thread.Message ID: ***@***.***>

-- Robert J. Harrison tel: 865-274-8544

[EricaCMitchell]

Since we are using the same input functions for Tensor<R> ovlp = matrix_inner(world, v, v);, the inner product of <v, v> would be real as a basic property of the inner product itself requires that with an arbitrary vector v, <v, v> is real and nonnegative.

[robertjharrison]

But these are vectors of functions. The diagonal of the matrix will be real, as you note, but the off diagonal elements can be complex.

…

On Tue, Oct 22, 2024 at 3:10 PM Erica Mitchell ***@***.***> wrote: Since we are using the same input functions for Tensor<R> ovlp = matrix_inner(world, v, v);, the inner product of <v, v> would be real as a basic property of the inner product itself requires that with an arbitrary vector v, <v, v> is real and nonnegative. I need to change this in more places in the code <https://github.com/m-a-d-n-e-s-s/madness/blob/8e4d98a5abcaab117740dc1c6778ba407686f955/src/madness/mra/vmra.h#L491C6-L491C49> as well, so wait on merge. — Reply to this email directly, view it on GitHub <#555 (comment)>, or unsubscribe <https://github.com/notifications/unsubscribe-auth/ABZSAPOW2J2DF42CMWHM3ETZ42PJHAVCNFSM6AAAAABQND4CAKVHI2DSMVQWIX3LMV43OSLTON2WKQ3PNVWWK3TUHMZDIMZQGA2DKNBYGI> . You are receiving this because you commented.Message ID: ***@***.***>

-- Robert J. Harrison tel: 865-274-8544

[EricaCMitchell]

This is starting to make sense to me. Thanks for the comments! I'm closing this.