The combination of Non-Local Correlation with other functionals

Questions
#1

Hi, is it possible to employ Non-Local Correlation (NLC) Functionals like VV10 in conjunction with B3LYP or TPSSh to account for dispersion effects? Or is it strictly limited to usage with PBE?
I mean something like this:

$molecule
0 1
Co                 8.44777400    5.67230600   12.57000400
P                  7.54495100    6.82940400   10.84548200
P                  9.38529200    4.87402900   14.45082300
O                  7.15492900    6.68316200   13.70969000
N                  7.09380800    4.11074000   12.39362700
C                  5.38233000    5.06406200   13.91195900
C                  4.11888300    4.76408000   14.49191600
H                  3.65738700    3.80119300   14.24897400
C                  7.35355600    2.93504400   11.62473400
C                  5.95072100    4.08225900   13.03340600
H                  5.31227000    3.19846000   12.87679600
C                  6.02026600    6.32574300   14.19932100
C                  4.09504700    6.88369100   15.62990700
H                  3.59853300    7.59351600   16.29862800
C                  5.32045700    7.21600000   15.07963000
H                  5.79729700    8.17400300   15.30137500
C                  3.47373400    5.64528700   15.34207600
H                  2.50570600    5.39349900   15.78008000
C                  9.87053600    4.79168900   11.49927800
H                  9.74553300    4.92130600   10.40956500
H                  9.91379800    3.70092200   11.67753800
H                 10.86679600    5.20225400   11.74192900
C                  7.49783800    0.50628600   11.44593800
H                  7.43371700   -0.47552100   11.92237200
C                  7.67408500    3.02818300   10.25993700
H                  7.74595800    4.01159600    9.79863200
C                  7.26858700    1.65635800   12.20864300
H                  7.03189700    1.56568100   13.27095300
C                  7.81005400    0.60965600   10.08740700
H                  7.99069200   -0.28815000    9.49152100
C                  9.70202600    7.21041700   12.76233900
H                 10.03596100    7.62030900   11.79014300
H                 10.63452500    6.96162900   13.30295100
H                  9.18853600    8.01060400   13.32324700
C                 11.16463900    4.39484900   14.41789300
H                 11.49935800    4.10214800   15.42617900
H                 11.78215700    5.23571700   14.06932800
H                 11.31348500    3.54814700   13.73186600
C                  7.89246800    1.87776900    9.49983500
H                  8.13113500    1.97458800    8.43752900
C                  9.33394700    6.06374100   15.85585800
H                  8.28842100    6.30378300   16.09532800
H                  9.84537300    6.99603400   15.57680400
H                  9.82328500    5.63506200   16.74541300
C                  8.64909300    3.37163500   15.23088100
H                  8.76448900    2.51116100   14.55558000
H                  7.57819800    3.53293500   15.42497300
H                  9.15164200    3.14227300   16.18480400
C                  8.42445100    6.92158600    9.22642600
H                  9.45959900    7.25954000    9.38682800
H                  7.91855200    7.62871400    8.54931800
H                  8.45889900    5.93432000    8.74102600
C                  7.29483800    8.61573300   11.22138000
H                  6.76489000    8.70431100   12.18025000
H                  6.70747600    9.10077100   10.42524900
H                  8.26496900    9.12441500   11.31187700
C                  5.83473300    6.36003000   10.32892300
H                  5.80510000    5.32518600    9.95783700
H                  5.46739900    7.03296200    9.53686800
H                  5.16092800    6.43489800   11.19592700
$end

$rem
BASIS                           def2-svp
METHOD                      TPSSh
NL_CORRELATION     vdW-DF-10
!......
$end
#2

Have you tried EXCHANGE = GEN in combination with a $xc_functional section?

#3

I think you can easily do this but the NL parameters are not optimized for these functionals and might be worse than optimized empirical dispersion correction.

#4

That may be true but I’m pretty sure we’ve done this, in order to implement the PBEh(40)-rVV10 functional that is used in some CP2K papers,
https://doi.org/10.1021/acs.jpclett.2c03460
https://doi.org/10.1021/acs.jctc.1c00412

I checked my old inputs from those papers and the NL correlation doesn’t go into the $xc_functional section but rather into $rem, e.g.

$rem
nl_correlation  vv10
use_rvv10       true
nl_vv_c         93
nl_vv_b         530
$end

where the last two lines are setting the parameters to some that have been used in CP2K calculations

#5

Thank you for your detailed explanation. Based on this, am I correct in using the following input?
I extracted the ‘b=5.0’ value for TPSS functional from https://doi.org/10.1021/ct4003527.

$rem
method            tpss
nl_correlation    vv10
use_rvv10         true
nl_vv_c           93
nl_vv_b           500
$end
#6

Please see the papers that I cited for the details about the functional, which you can map to the input section that I pasted above. These keywords should also be described in the manual.