Possibly incorrect GTH-PBE pseudopotential for Tl

[Nicholas Winner]

I was optimizing Tl lattice parameter, and found a large discrepancy with published PBE results.  On a hunch, I decided to re-optimize the pseudopotential, and based on what I tried, it seems the GTH pseudopotential for Tl that ships with CP2K is not correct. The published potential is:

```
Tl GTH-PBE-q13 GTH-PBE
    2    1   10
     0.55000000    1    12.29058515
    3
     0.51280306    2     7.19017017    -2.86301073
                                        3.69613096
     0.57711505    2     4.76094580    -3.67332496
                                        4.34633671
     0.39323001    2   -11.01268700     6.42159202
                                       -7.28140093
```

But, if you run pseudopotential optimization, then you get the following. Note only 1 number changes: `0.39323001  -> 0.32379979 `

```
Tl GTH-PBE-q13
    2    1   10    0
    0.55000000       1   12.29058515
       3
    0.51280306       2    7.19017017   -2.86301073
                                        3.69613096
    0.57711505        2    4.76094580  -3.67332496
                                        4.34633671
    0.32379979       2  -11.01268700    6.42159202
                                       -7.28140093
```

The input file I used to generate this is is listed at the end of this message. 

Can the devs, especially Prof. Hutter, comment on whether this is a real issue? If it is, then I hope it can be updated in the repo.

&GLOBAL

        PROGRAM_NAME ATOM

&END GLOBAL

&ATOM

        ELEMENT Tl

        RUN_TYPE PSEUDOPOTENTIAL_OPTIMIZATION

        ELECTRON_CONFIGURATION [Xe] 6s2 5d10 6p1

        CORE [Xe]

        ! MAX_ANGULAR_MOMENTUM 2

        COULOMB_INTEGRALS ANALYTIC

        EXCHANGE_INTEGRALS ANALYTIC

        &METHOD

                METHOD_TYPE KOHN-SHAM

                RELATIVISTIC DKH(3)

                &XC

                        &XC_FUNCTIONAL PBE

                        &END XC_FUNCTIONAL

                &END XC

        &END METHOD

        &OPTIMIZATION

                EPS_SCF 5e-9

                max_iter 1000

                damping .2

                n_diis 15

        &END OPTIMIZATION

        &PRINT

                &BASIS_SET

                &END

        &END

        &AE_BASIS

                BASIS_TYPE GEOMETRICAL_GTO

        &END AE_BASIS

        &PP_BASIS

                BASIS_TYPE GEOMETRICAL_GTO

        &END PP_BASIS

        &POTENTIAL

                PSEUDO_TYPE GTH

                POTENTIAL_FILE_NAME POTENTIAL

                POTENTIAL_NAME GTH-PBE-q13

        &END POTENTIAL

        &POWELL

                ACCURACY 1.e-10

                STEP_SIZE 0.5

                WEIGHT_PSIR0 0.1

        &END POWELL

&END ATOM

[Jürg Hutter]

Hi

your input for the PP optimization misses the 4f electrons

ELECTRON_CONFIGURATION [Xe] 4f14 6s2 5d10 6p1
CORE [Xe] 4f14

I don't think the optimization will change the PP parameters significantly.

In POTENTIAL_UZH and BASIS_MOLOPT_UZH you can find alternative PP and Basis sets.
You could for example try q21.

regards
JH

________________________________________
From: cp...@googlegroups.com <cp...@googlegroups.com> on behalf of Nicholas Winner <nwi...@berkeley.edu>
Sent: Wednesday, October 26, 2022 5:07 AM
To: cp2k
Subject: [CP2K:17935] Possibly incorrect GTH-PBE pseudopotential for Tl

--
You received this message because you are subscribed to the Google Groups "cp2k" group.
To unsubscribe from this group and stop receiving emails from it, send an email to cp2k+uns...@googlegroups.com<mailto:cp2k+uns...@googlegroups.com>.
To view this discussion on the web visit https://groups.google.com/d/msgid/cp2k/ca5634d0-1e75-41b6-8821-77b4b0ba76e5n%40googlegroups.com<https://groups.google.com/d/msgid/cp2k/ca5634d0-1e75-41b6-8821-77b4b0ba76e5n%40googlegroups.com?utm_medium=email&utm_source=footer>.

[Nicholas Winner]

Thanks professor, I see the mistake now. I'll retry with POTENTIAL_UZH and see if it gives improvements. I was doing some benchmarking and found Tl compounds had very poor property predictions, maybe q21 will resolve that.

[Nicholas Winner]

I see no basis sets for 21 electron configuration of Tl in the data directory. Are those available anywhere?

[Jürg Hutter]

Check BASIS_MOLOPT_UZH, there you can find

SVP-MOLOPT-PBE-GTH-q21
TZVPP-MOLOPT-PBE-GTH-q21
QZVPP-MOLOPT-PBE-GTH-q21

regards

JH

________________________________________
From: cp...@googlegroups.com <cp...@googlegroups.com> on behalf of Nicholas Winner <nwi...@berkeley.edu>

Sent: Thursday, October 27, 2022 10:56 PM
To: cp2k
Subject: Re: [CP2K:17946] Possibly incorrect GTH-PBE pseudopotential for Tl

To view this discussion on the web visit https://groups.google.com/d/msgid/cp2k/5e557ed1-4e5a-49d1-bf57-1d7189c6c179n%40googlegroups.com<https://groups.google.com/d/msgid/cp2k/5e557ed1-4e5a-49d1-bf57-1d7189c6c179n%40googlegroups.com?utm_medium=email&utm_source=footer>.

[Nicholas Winner]

I see now, v2022.1 did not have them, but I can find them in the repo.

Thanks for the help professor.