Numerical gradients per region

The numerical gradients that can be used within QUILD are now completely generalized, and can be specified per region. For instance, if one wants to use the numerical gradient of a hybrid-metagga functional (such as e.g. mPBE0KCIS) for a small part of the total system, and the analytical gradient of a GGA for the rest, it is simple and straightforward to achieve this:

QUILD   
  SMETAGGA mPBE0KCIS
  NR_REGIONS=2
        
  INTERACTIONS
    TOTAL     description 1
    REPLACE region 1   description 3 for description 2
  SUBEND
  
  REGION 1
    1-3
  SUBEND
  REGION 2
    4-6
  SUBEND
  DESCRIPTION 1 NEWMM
  ! MM input for NewMM, not explicitly shown here
  SUBEND

  DESCRIPTION 2 NEWMM
  ! MM input for NewMM, not explicitly shown here
  SUBEND

  DESCRIPTION 3 NUMGRAD MGGA
    BASIS
     type DZ
     core small
    END
    SCF
     Converge 1.0e-7 1.0e-7
     Iterations 99
     diis ok=0.01
    END
    INTEGRATION 7.0 7.0 7.0
    CHARGE   0.0
    METAGGA
  SUBEND

END