From owner-chemistry@ccl.net Tue May 31 15:55:01 2022 From: "Henrique C. S. Junior henriquecsj*o*gmail.com" To: CCL Subject: CCL: Magnetism and luminescence for large lanthanide clusters Message-Id: <-54718-220531142356-21351-RhFH3EqNlvZQBgR1s3OyuQ[-]server.ccl.net> X-Original-From: "Henrique C. S. Junior" Content-Type: multipart/alternative; boundary="0000000000009f474c05e052df28" Date: Tue, 31 May 2022 15:22:49 -0300 MIME-Version: 1.0 Sent to CCL by: "Henrique C. S. Junior" [henriquecsj+/-gmail.com] --0000000000009f474c05e052df28 Content-Type: text/plain; charset="UTF-8" Content-Transfer-Encoding: quoted-printable Dear professors Neese and Riplinger, thank you for your time and kind reply= . I am, indeed, using the latest ORCA version and it still baffles me how fast it can be. Because I am using the 5.0.3 version, I am quite confident that even the default settings are a good way to get nice results. My group has access to single-crystal X-Ray Diffraction and to a cryogenic magnetometer, so we have (usually) a good experimental starting point. At most I usually reoptimize the hydrogens. This new cluster series, as I mentioned, have 15 to 20 lanthanides from Sm to Dy and the molecules are =E2=80=9CColosseum-shaped=E2=80=9D with C2 symmetry. I tried to optimize the hydrogens using one of the compound methods (r2SCAN-3c, PBE-3c) but even this task is taking days for a single step in the relaxation. Using the TRAH converge or KDISS only, tweaking the settings, was still taking too long on a cluster with 64 GB per node and 48 cores. My initial idea was to produce at least a good spin description. Ideally, I would like to get magnetization and susceptibility, SOC, anisotropy parameters, absorption, and emission spectra. At this point I have used XTB2 and the orca %compound language to evaluate every possible multiplicity to find, roughly, the most stable. After that I am running (for almost five hundred hours) a single-point calculation with SARC2-DKH-QZVP for the lanthanides and DKH-Def2-SVP for the lighter elements testing TPSSh and r2SCAN. The total number of basis functions is 5130 and 14375 for AuxJ. My single-point using r2SCAN-3c finished after ~ 2 days, but the resulting spin density was kind of strange, so I=E2=80=99m lo= oking for a confirmation method. Because the system has proven to be so difficult, I was losing hope about using CASSCF or any DLPNO feature but I will definitely look at the multiscale methods available and the hints that you provided. Thank you so much. On Mon, May 30, 2022 at 2:32 PM Christoph Riplinger riplinger^faccts.de < owner-chemistry(_)ccl.net> wrote: > > Sent to CCL by: Christoph Riplinger [riplinger-#-faccts.de] > Dear Henrique, > > great to hear that ORCA has proven useful for your single-molecule > magnet research already. I think you should be able to use ORCA also for > your new use case. Maybe you have to use some features that you have not > yet been aware of before. Some suggestions / questions: > > * Are you already using ORCA 5? If not, I strongly suggest to upgrade to > ORCA 5.0.3 - which is way more efficient and robust in terms of speed > and accuracy compared to previous versions. > > * Why is it taking too long? Is the calculation not converging, or is > each cycle just too slow? Is our TRAH solver activated in your > calculations? (We tested it for metallic clusters for which it worked > well. If you still run into convergence problems, best to post the > details in the ORCA forum.) > > * Are you using pseudopotentials already (at least for geometry > optimization)? They are implemented for lanthanides (and heavier > elements). They are already switched on by default for the def2 basis > sets. For other basis sets it depends. > > * You could use non-hybrid DFT as a guess for a CASSCF calculation. > > If your system is too large (depending on the method you are using, that > is 10k-20k basis functions), maybe have a look at the following: > > * You could use ORCA's new multiscale features. I could imagine that > ORCA's QM/R2SCAN-3C, or, if the system is much too large, QM/XTB, could > be useful for your use case (both available also for lanthanides). > Please have a look at our multiscale tutorials: > https://www.orcasoftware.de/tutorials_orca/multi/basics.html . > > * You could use smaller basis sets or larger ECPs for less important > parts of the molecule. Do you already know about ORCA's new FragBasis > capabilities? You can assign basis sets, ECPs, etc. to entire fragments > (see "Assigning Basis Sets and ECPs to Fragments" in the ORCA manual) > which can be conveniently defined as described in "Fragment > Specification" in the ORCA manual. > > * For DLPNO-CCSD(T) calculations you can use the DLPNO-multilevel > implementation (which can also be combined with the multiscale feature). > > * If you are using DLPNO-CCSD(T) or CASSCF, you have to switch on > RIJCOSX explicitly by using !RIJCOSX . > > * With respect to magnetism / SOC you should also have a look at our SOC > tutorial - which is written for TDDFT, but you can also apply the same > to CASSCF: https://www.orcasoftware.de/tutorials_orca/spec/SOC.html. > > Hope that helps, > > Christoph > > On 29.05.22 15:15, Henrique Castro henriquecsj ~~ outlook.com wrote: > > Sent to CCL by: "Henrique Castro" [henriquecsj]*[outlook.com] > > During my masters and my PhD, I was working mostly with Single-Ion > Magnets and > > small Single-Molecule Magnets (at most 2 transition metals or > lanthanides). > > With such systems I was able to calculate broken-symmetries, > CASSCF+NEVPT2, > > CCSD(T) with DLPNO and ask for some pretty fancy stuff that the ORCA > software > > makes possible at a small cost. > > > > Now the group is exploring new areas and the first large lanthanide > clusters > > are in the queue to see if we can calculate something interesting. The > new > > systems are much larger than what Ive dealt with before (> 255 atoms) > with > > molecules containing 15 and even 20 lanthanides (Eu, Sm, Gd, Dy). Even = a > > single-point is taking weeks. > > > > So, Id like to ask for your advice on how it is possible (if possible a= t > all) > > to deal with such systems in a reasonable computational level for > magnetism and > > luminescence. Is there a way to consider relativistic effects and use > > multireference methods with such monsters? Maybe this is something that > I cant > > reach using ORCA but is there a software that can deal with it? > > > > Thanks for your comments.> > > > -- > FAccTs GmbH > Rolandstrasse 67, 50677 K=C3=B6ln > Amtsgericht K=C3=B6ln HRB 88406 > Gesch=C3=A4ftsf=C3=BChrer: Dr. Christoph Riplinger > https://www.faccts.de | https://twitter.com/faccts_orca > > > > -=3D This is automatically added to each message by the mailing script = =3D-> > > --=20 *Henrique C. S. Junior* --0000000000009f474c05e052df28 Content-Type: text/html; charset="UTF-8" Content-Transfer-Encoding: quoted-printable

Dear professors Neese and Riplinger, thank you for your time and kind reply.

I am, indeed, us= ing the latest ORCA version and it still baffles me how fast it can be. Because I am using= the 5.0.3 version, I am quite confident that even the default settings are a go= od way to get nice results.

My group has acc= ess to single-crystal X-Ray Diffraction and to a cryogenic magnetometer, so we have (usually) a g= ood experimental starting point. At most I usually reoptimize the hydrogens. Th= is new cluster series, as I mentioned, have 15 to 20 lanthanides from Sm to Dy= and the molecules are =E2=80=9CColosseum-shaped=E2=80=9D with C2 symmetry.

I tried to optim= ize the hydrogens using one of the compound methods (r2SCAN-3c, PBE-3c) but even this task is taking days for a single step in the relaxation. Using the TRAH converge or KDISS only, tweaking the settings, was still taking too long on a cluster w= ith 64 GB per node and 48 cores.

My initial idea = was to produce at least a good spin description. Ideally, I would like to get magnetization a= nd susceptibility, SOC, anisotropy parameters, absorption, and emission spectr= a.

At this point I = have used XTB2 and the orca %compound language to evaluate every possible multiplicity to = find, roughly, the most stable. After that I am running (for almost five hundred hours) a single-point calculation with SARC2-DKH-QZVP for the lanthanides a= nd DKH-Def2-SVP for the lighter elements testing TPSSh and r2SCAN. The total number of basi= s functions is 5130 and 14375 for AuxJ. My single-point using r2SCAN-3c finis= hed after ~ 2 days, but the resulting spin density was kind of strange, so I=E2= =80=99m looking for a confirmation method.

Because the syst= em has proven to be so difficult, I was losing hope about using CASSCF or any DLPNO feature = but I will definitely look at the multiscale methods available and the hints th= at you provided.

=C2=A0

Thank you so muc= h.


On Mon, May 30, 2022 at 2:32 PM Christoph Riplinger riplinger^<= a href=3D"http://faccts.de">faccts.de <owner-chemistry(_)ccl.net> wrote:

Sent to CCL by: Christoph Riplinger [riplinger-#-faccts.de]
Dear Henrique,

great to hear that ORCA has proven useful for your single-molecule
magnet research already. I think you should be able to use ORCA also for your new use case. Maybe you have to use some features that you have not yet been aware of before. Some suggestions / questions:

* Are you already using ORCA 5? If not, I strongly suggest to upgrade to ORCA 5.0.3 - which is way more efficient and robust in terms of speed
and accuracy compared to previous versions.

* Why is it taking too long? Is the calculation not converging, or is
each cycle just too slow? Is our TRAH solver activated in your
calculations? (We tested it for metallic clusters for which it worked
well. If you still run into convergence problems, best to post the
details in the ORCA forum.)

* Are you using pseudopotentials already (at least for geometry
optimization)? They are implemented for lanthanides (and heavier
elements). They are already switched on by default for the def2 basis
sets. For other basis sets it depends.

* You could use non-hybrid DFT as a guess for a CASSCF calculation.

If your system is too large (depending on the method you are using, that is 10k-20k basis functions), maybe have a look at the following:

* You could use ORCA's new multiscale features. I could imagine that ORCA's QM/R2SCAN-3C, or, if the system is much too large, QM/XTB, could=
be useful for your use case (both available also for lanthanides).
Please have a look at our multiscale tutorials:
https://www.orcasoftware.de/tutorials_or= ca/multi/basics.html .

* You could use smaller basis sets or larger ECPs for less important
parts of the molecule. Do you already know about ORCA's new FragBasis <= br> capabilities? You can assign basis sets, ECPs, etc. to entire fragments (see "Assigning Basis Sets and ECPs to Fragments" in the ORCA man= ual)
which can be conveniently defined as described in "Fragment
Specification" in the ORCA manual.

* For DLPNO-CCSD(T) calculations you can use the DLPNO-multilevel
implementation (which can also be combined with the multiscale feature).
* If you are using DLPNO-CCSD(T) or CASSCF, you have to switch on
RIJCOSX explicitly by using !RIJCOSX .

* With respect to magnetism / SOC you should also have a look at our SOC tutorial - which is written for TDDFT, but you can also apply the same
to CASSCF: https://www.orcasoftware.de/tutor= ials_orca/spec/SOC.html.

Hope that helps,

Christoph

On 29.05.22 15:15, Henrique Castro henriquecsj ~~ outlook.com wrote:
> Sent to CCL by: "Henrique=C2=A0 Castro" [henriquecsj]*[outlook.com<= /a>]
> During my masters and my PhD, I was working mostly with Single-Ion Mag= nets and
> small Single-Molecule Magnets (at most 2 transition metals or lanthani= des).
> With such systems I was able to calculate broken-symmetries, CASSCF+NE= VPT2,
> CCSD(T) with DLPNO and ask for some pretty fancy stuff that the ORCA s= oftware
> makes possible at a small cost.
>
> Now the group is exploring new areas and the first large lanthanide cl= usters
> are in the queue to see if we can calculate something interesting. The= new
> systems are much larger than what Ive dealt with before (> 255 atom= s) with
> molecules containing 15 and even 20 lanthanides (Eu, Sm, Gd, Dy). Even= a
> single-point is taking weeks.
>
> So, Id like to ask for your advice on how it is possible (if possible = at all)
> to deal with such systems in a reasonable computational level for magn= etism and
> luminescence. Is there a way to consider relativistic effects and use<= br> > multireference methods with such monsters? Maybe this is something tha= t I cant
> reach using ORCA but is there a software that can deal with it?
>
> Thanks for your comments.>
>
--
FAccTs GmbH
Rolandstrasse 67, 50677 K=C3=B6ln
Amtsgericht K=C3=B6ln HRB 88406
Gesch=C3=A4ftsf=C3=BChrer: Dr. Christoph Riplinger
http= s://www.faccts.de | https://twitter.com/faccts_orca



-=3D This is automatically added to each message by the mailing script =3D-=
E-mail to subscribers: CHEMISTRY(_)ccl.net or use:
=C2=A0 =C2=A0 =C2=A0 http://www.ccl.net/cgi-bin/ccl/s= end_ccl_message

E-mail to administrators: CHEMISTRY-REQUEST(_)ccl.net or use
=C2=A0 =C2=A0 =C2=A0 http://www.ccl.net/cgi-bin/ccl/s= end_ccl_message
=C2=A0 =C2=A0 =C2=A0 http://www.ccl.net/chemistry/sub_un= sub.shtml

Before posting, check wait time at: http://www.ccl.net

Job: http://www.ccl.net/jobs
Conferences: http://server.ccl.net/chemist= ry/announcements/conferences/

Search Messages: http://www.ccl.net/chemistry/sear= chccl/index.shtml
=C2=A0 =C2=A0 =C2=A0 http://www.ccl.net/spammers.txt

RTFI: http://www.ccl.net/chemistry/aboutccl/ins= tructions/




--
= Henrique C. S. Junior

--0000000000009f474c05e052df28--