Why can't I use LDA with DFT-D?

[395235863]

Thank you in advance.

[Anders Blom]

Ask the people who wrote the original article about DFT-D
The method is simply designed for GGA (that is, the parameters are "fitted" or chosen for GGA). Note that it's an empirical approach!

[Jess Wellendorff]

The reason the DFT-D approach was not designed for use with LDA is most likely that LDA already gives "vdW-like" binding in itself, spurious as it may be.

[Anders Blom]

The right results for the wrong reasons

[395235863]

http://i1.rgstatic.net/figure/265477046_fig3_Figure-23Comparison-of-calculated-(DFT-D2-with-LDA)-and-experimentally-measured-(ARPES)/AS:271626805706756@1441772285161。

here i get the pic of 
Comparison of calculated (DFT-D2 with LDA) and experimentally measured (ARPES)

[zh]

http://i1.rgstatic.net/figure/265477046_fig3_Figure-23Comparison-of-calculated-(DFT-D2-with-LDA)-and-experimentally-measured-(ARPES)/AS:271626805706756@1441772285161。

here i get the pic of 
Comparison of calculated (DFT-D2 with LDA) and experimentally measured (ARPES)

I guess it may be a typo for "DFT-D2 with LDA" on the figure. 

[Anders Blom]

Might be that someone made such a calculation, but it doesn't mean it's a good idea... As mentioned above, sometimes you get the right results for the wrong reasons.

[zh]

The issue is that the magic values of parameters in DFT-D2 are available in  Grimme's paper and implemented in ATK only for the GGA-type functional. How did the author of "Comparison of calculated (DFT-D2 with LDA) and experimentally measured (ARPES)" perform the  DF2-D2 calculations with LDA?
Even they are very lucky to get reasonable results in agreement with experiments for their system, but this doesn't make sense. 

The LDA usually gives a stronger bonding than the real one. The DFT-D2 with LDA would give much stronger bonding than LDA. So the use of DFT-D2 with LDA may be quite questionable.