Comments obtained over the years

2025

• would be nice to know which ones were in libxc.
• It’d be nice to have a poll like this for Gaussian basis sets or planewave pseudopotentials!
• “Hate” is a bit strong
• wB9X-3c (I realize this is a refit wB97X-D4/vDZP, but it is my #2 functional after r²SCAN-3c)
• https://www.physics.rutgers.edu/gbrv/ PP are good for phonon calculations.
• LH20t, wlh22t, lh24n, wdh25
• NA
• adding a dispersion correction, especially a non-self-consistent one, doesn’t make it a new functional
• I have filled the like or neutral, for the functional which i have used i have given liked or for the functional which i have not used i have given neutral.
• mBJ for band gap calculations
• HSE06!!!!! the best functional in the world!
• Maybe include a option “I have never used this functional”
• It would be good to include RSCAN. SCAN and R2SCAN are good in theory, but impractical in real calculations because they are ill-conditioned in low density regions. The only way to use them even remotely reliably is to force your basis set to go to zero faster than the instability, which in practice usually means you have to use GTOs. There are actually two instabilities; R2SCAN fixes the easier one, which can make some calculations stable enough, but it is still too unreliable.
• R2scan0-D4 and wB97X-3c would be nice
• I hate DFT in general
• For th functionals that do not have a specific dispersion-corrected re-parametrisation there should not be two versions (B3LYP/B3LYP-D), you should always use dispersion correction. I know it's in there for historical reasons, but I assume more people just vote the same way for both.
• Do scoring out of 5 for each instead of out of 3

2024

• [Functional] + [XDM, TS, MBD]
• For transition metal, I like B3LYP and BP86 and for main group chemistry, M062X
• None of the above = I haven't used it before.
• B86bPBE
• As evidenced by the references, e.g., "B3LYP-D" is not really a separate functional, but simply B3LYP, together with one specific version of Grimme's dispersion correction. A dispersion correction should always be used, but whether this specific one, or a more recent version like D4, or a non-local correlation functional like VV10, is a separate question. Similarly, Chai and Head-Gordon's wB97X-D refit is not particularly impressive, but the wB97X-V functional with a drop-in replacement of VV10 by D3(BJ) or D4 is rather good. Lumping nuanced factors like that into a single (Gaussian-flavored) keyword is likely to skew the results of the poll.
• Missing the mention of MBD
• I am neutral on all functionals that did not test myself. But I cannot say I really enjoy a particular functional since there's no exact one at all.
• B97D3
• Include TPSSh
• Few new ones in there for me to go explore. Thanks.

2022

• "I would like more differentiation maybe scale of 1-5, 1 is hate, 5 is like"
• "Several interesting DFs missing including various DSD-, wB97X-2, MN15, M08HX, while some rather unusual were included e.g. DM21, S12g. Why?"
• "Density-functional theory, formerly an ab initio method, has successfully turned semi-empirical, let's put it this way."
• "I have used very few of the functional listed here. So, I used none of the above for most of the answers."

2021

• I do not use those "new suggestions" so cannot comment on them
• Hate almost all of it

2020

• Like MN15L
• It's a little hard to rank a functional with and without -D. I feel like with the ubiquity of dispersion corrections these days, it might be worth voting under the assumption that a D correction can be applied if desired.
• SCAN-rVV10 should be in the list.
• In my experience PBE0 give better results with similar computational cost as HSE06 and others are sophisticated but too much in demand for computer power (of course LDA, PBE are still very cheap)
• SCAN is very hard to converge with slab + vacuum!!!!!!!!
• I consistently find that CAM-B3LYP is good for TD-DFT of transition metal complexes, M06-2X for main group mechanistic studies, B3LYP good for organic molecules (good = most replicates experimental data).
• There no reason for hate
• PBE0-DH best of the double hybrids
• New favourite = rSCAN - not in your list?
• What about SCAN or optb86b and vdW-DF functionals?

2018

• PBEsol?
• WHERE IS BEEF-VDW???

2017

• TPSSh is 3good5you -> Don Truhlar
• I used just two types of functionals, B3LYP and B3PW91, so I polled these types only.

2016

• No MCY
• tl;dr
• Do some real chemistry like Heisenberg in breaking bad, make some blue meth. Don't waste precious taxpayers money by just playing on your computers.

2015

• The M11 series of functionals should be included !! (I have no link with Minnesota...)
• LDA: THE ONE TRUE FUNCTIONAL. ALL OTHERS SHALL PERISH.
• Please add BEEF-vdW as functional. Survey is clumsy to use, i.e. why can't one add AND vote for new functionals. And the presentation of results at http://www.marcelswart.eu/ is really opaque. Too bad since the effort is really useful.
• Why are vdW-DF functionals of Langreth and Lundqvist not included in the poll?